The research, published in BMC Medical Imaging, tackles one of the most consequential decisions in urologic oncology: predicting the International Society of Urological Pathology, or ISUP, grade of prostate cancer before surgery. The ISUP grade, derived from the Gleason scoring system, describes how abnormal cancer cells look under a microscope and is one of the strongest predictors of how a tumor will behave. Low-grade tumors can often be monitored safely under active surveillance, while high-grade disease may demand radical prostatectomy, radiation, or systemic therapy. Getting that grade right before treatment — rather than after the gland is removed — could spare thousands of men from unnecessary surgery while ensuring that dangerous cancers are not undertreated.

The challenge is that the standard imaging tool, multiparametric magnetic resonance imaging, or mpMRI, has well-known blind spots. mpMRI excels at revealing anatomy: it shows the size and location of suspicious lesions, the integrity of the surrounding capsule, and features such as restricted water diffusion that hint at cellular density. But it offers only limited insight into tumor metabolism and molecular biology. Some biologically aggressive lesions — so-called occult high-risk disease — simply do not light up convincingly on an MRI, evading detection until pathology delivers an unwelcome surprise after surgery.

That is where the second modality comes in. The study used 18F-PSMA-1007 positron emission tomography combined with computed tomography, a molecular imaging technique that exploits a quirk of prostate cancer biology: most prostate cancer cells overexpress a protein called prostate-specific membrane antigen, or PSMA. The radiotracer 18F-PSMA-1007 binds to this protein, allowing the PET scanner to map where PSMA-hungry tumor cells are concentrated throughout the gland and the body. Because uptake intensity often correlates with tumor aggressiveness, PET/CT provides a molecular readout that MRI lacks. The trade-offs are real — higher cost and a dose of ionizing radiation — which is precisely why the researchers wanted to know whether combining the two scans with machine learning genuinely adds predictive power, rather than simply adding expense.

To answer that question, the team assembled a retrospective cohort of 341 patients who had undergone preoperative mpMRI, 18F-PSMA-1007 PET/CT, and radical prostatectomy at their institutions, along with an independent external validation cohort of 36 patients from a different center. The reference standard was uncompromising: the final ISUP grade assigned by pathologists to each surgically removed prostate. Against that gold standard, the researchers built a family of machine learning models and compared them head to head. There were five binary models — predicting simply whether a tumor was clinically significant or not — based respectively on clinical variables alone, on mpMRI alone, on PET/CT alone, on the fused imaging pair, and on the fused imaging pair plus clinical features. They also built three-class versions of the fusion models to predict full ISUP grade groups, a harder task that mirrors the granularity clinicians actually need.

The results tell a striking story about the power of multimodal fusion. In the internal cohort, a model using clinical variables alone achieved an area under the receiver operating characteristic curve, or AUC, of 0.739 — modest performance, as one would expect from variables such as age, PSA levels, and biopsy findings. The mpMRI model reached an AUC of 0.881, and the PET/CT model nearly matched it at 0.888. But when the two imaging streams were fused, performance jumped: the combined mpMRI plus PET/CT model achieved an AUC of 0.945, and adding clinical features nudged it to 0.950. In practical terms, that difference moves the model from useful to potentially transformative, because AUC values above 0.9 indicate excellent discrimination between aggressive and indolent disease — approaching the kind of reliability that could meaningfully inform treatment decisions.

The three-class task, which asks the model to distinguish low, intermediate, and high-grade disease rather than drawing a single binary line, proved predictably harder. The imaging fusion model achieved a macro-AUC of 0.810 with an accuracy of 0.695, rising to 0.830 macro-AUC and 0.698 accuracy when clinical features were added. Those numbers reflect genuine, though imperfect, discrimination across three categories — respectable for a problem where even expert pathologists sometimes disagree, but a reminder that predicting a pathology grade from pixels remains fundamentally harder than splitting tumors into two groups.

What sets this study apart from much of the medical AI literature is its insistence on interpretability. Deep learning models are notorious black boxes, and radiologists are understandably reluctant to trust a prediction they cannot interrogate. The researchers deployed two complementary explanation techniques. SHAP, which stands for SHapley Additive exPlanations, quantified how much each clinical feature contributed to each prediction, borrowing a concept from game theory to divide credit fairly among inputs. Grad-CAM, short for Gradient-weighted Class Activation Mapping, generated heatmaps overlaid on the medical images, revealing exactly which regions of the prostate the neural network looked at when making its decision. If the model flags a lesion that the radiologist can see and evaluate — rather than keying in on an artifact or a scanner-specific signature — clinicians have a concrete reason to trust it, and a concrete reason to be skeptical when the highlighted regions make no anatomical sense.

The external validation results temper the enthusiasm appropriately, and the authors are candid about it. In the 36-patient independent cohort from a different center, the fused models achieved binary AUCs of 0.726 and 0.718 — a noticeable drop from the 0.945 and 0.950 seen internally — while three-class macro-AUCs held up somewhat better at 0.778 and 0.808. This kind of performance decay across institutions is one of the most common failure modes in medical imaging AI: models can inadvertently learn center-specific scanner settings, imaging protocols, or patient populations rather than generalizable biology. The authors explicitly caution that the external validation is preliminary and that the small external cohort warrants careful interpretation. It is a refreshingly honest framing in a field where inflated claims have sometimes outrun the evidence.

Where does this leave patients and their doctors? The study, registered prospectively at the Chinese Clinical Trial Registry in October 2024, positions the fused model not as a replacement for pathology — the authors emphasize that it supplements histopathology rather than supplanting it — but as a decision-support adjunct. In a plausible clinical workflow of the future, a man with a suspicious MRI and an indeterminate biopsy might undergo PSMA PET/CT as part of his standard workup, and an interpretable fusion model would combine all available evidence into a calibrated estimate of tumor aggressiveness. That estimate could help determine whether active surveillance is genuinely safe, whether a targeted biopsy should be repeated, or whether definitive treatment should proceed without delay. For men with occult high-grade disease hiding from conventional MRI, the molecular signal from PSMA imaging, interpreted by an algorithm trained on hundreds of surgically confirmed cases, could be the difference between catching an aggressive cancer early and discovering it too late.

The road to that future runs through prospective, multicenter validation, and the authors say as much. Larger and more diverse external cohorts will be needed to confirm that the performance gains survive contact with different scanners, protocols, and populations, and the modest external results suggest that domain generalization remains the central technical hurdle. Still, the study offers a compelling proof of concept that anatomical, functional, and molecular information are genuinely complementary — that the whole of two imaging modalities, intelligently fused and transparently explained, exceeds the sum of its parts. As prostate cancer affects millions of men worldwide, an AI that can predict tumor aggressiveness before surgery, and show radiologists exactly why it reached its conclusion, represents a meaningful step toward precision medicine in one of the most common cancers on Earth.

Subject of Research: Preoperative prediction of ISUP grade in prostate cancer using interpretable multimodal fusion of 18F-PSMA-1007 PET/CT and mpMRI with machine learning

Article Title: Interpretable multimodal fusion of 18F-PSMA-1007 PET/CT and mpMRI for preoperative ISUP grade prediction in primary prostate cancer

Article References: Zhong, J., Zhou, Y., Cheng, W., Wu, T., Yuan, Y., Yao, F., Zhuang, Y., Lin, Q., Li, T., Yang, Y., Lin, Y., & Ye, Y. (2026). Interpretable multimodal fusion of 18F-PSMA-1007 PET/CT and mpMRI for preoperative ISUP grade prediction in primary prostate cancer. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02752-y

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02752-y

Keywords: prostate cancer, ISUP grade, 18F-PSMA-1007 PET/CT, multiparametric MRI, multimodal fusion, machine learning, deep learning, SHAP, Grad-CAM, medical imaging, radiomics, decision support

The central innovation of the review lies in how it framed its primary outcome. Rather than asking whether better scans improve overall diagnostic accuracy, the study standardized the outcome as the proportion of MRI-positive units, whether patients, lesions, or regions, that were subsequently confirmed to harbor clinically significant prostate cancer, abbreviated csPCa, on histopathology. This distinction matters enormously. Diagnostic accuracy metrics such as sensitivity and specificity depend on the full spectrum of examined patients, including those with negative scans, whereas the confirmation rate among MRI-positive findings speaks directly to the practical question that radiologists and urologists face every day: when a scan flags a suspicious area, how likely is that flag to correspond to genuinely dangerous cancer?

The technical core of the review is its careful handling of the unit of analysis, a methodological issue that has plagued the imaging literature for years. Studies reporting results at the patient level, the region level, and the lesion level are statistically non-interchangeable, because observations within the same patient are correlated and cannot simply be pooled as if they were independent. Recognizing this, the author refused to combine the extractable comparative data, which consisted of one patient-level study, one region-level study, and one lesion-level study, into a single meta-analytic estimate. Instead, findings were synthesized separately by unit of analysis, and risk of bias was assessed using QUADAS-2, the standard tool for evaluating the methodological quality of diagnostic accuracy studies. This conservative approach, while it limits the statistical power of the conclusions, protects readers from the false precision that arises when correlated observations are treated as independent data points.

The headline numbers from the three studies with extractable comparative counts are strikingly divergent. At the patient level, higher-quality MRI was associated with csPCa confirmation in 48.0 percent of MRI-positive patients, compared with 35.3 percent among lower-quality scans, yielding an unadjusted relative risk of 1.36 with a 95 percent confidence interval of 0.92 to 2.02, an interval that crosses the null value of one and therefore does not reach conventional statistical significance. At the region level, the corresponding figures were 56.1 percent versus 36.2 percent, a descriptive unadjusted relative risk of 1.55, suggesting a potentially meaningful advantage for higher-quality imaging when the analysis is anchored to anatomical zones rather than whole patients. Yet at the lesion level, the direction reversed entirely: 45.4 percent versus 48.0 percent, a descriptive relative risk of 0.95, implying essentially no benefit, and perhaps a trivial disadvantage, for higher-quality scans when individual suspicious lesions are the unit of comparison.

That reversal across units of analysis is the most intellectually provocative finding of the review, and it deserves careful interpretation. One plausible explanation is that patient-level analyses capture the cumulative benefit of image quality across the entire gland, including the detection of cancers that would otherwise be missed altogether, whereas lesion-level analyses condition on a suspicious finding already being present, thereby restricting the comparison to lesions visible under both quality conditions. In other words, a better scan may help radiologists find cancers that a poorer scan never flags at all, and this detection benefit is invisible when the analysis is restricted to lesions that both scans identified. Cluster-aware confidence intervals could not be derived for the region-level or lesion-level estimates because the underlying studies did not report the information needed to account for within-patient clustering, which means the precision of those descriptive ratios remains unknown.

Beyond the three quantitative comparisons, the wider evidence base included in the review painted a directionally inconsistent picture, with results shaped by selection bias, verification bias, and concerns about applicability. Selection bias arises when the population undergoing MRI is not representative of the clinical population at large, for example when only patients with elevated prostate-specific antigen levels or prior negative biopsies are imaged. Verification bias, arguably the most pernicious threat in this field, occurs when only MRI-positive patients undergo biopsy, leaving MRI-negative patients without a histopathological reference standard and thereby inflating apparent accuracy. Applicability concerns include differences in scanner hardware, field strength, acquisition protocols, and the scoring systems used to grade image quality, most notably PI-QUAL, the five-point quality score developed to standardize the assessment of prostate MRI examinations, alongside PI-RADS, the structured reporting system for suspicion of clinically significant cancer.

The review’s bottom-line conclusion is deliberately restrained. The available evidence does not establish that higher MRI quality improves overall PI-RADS diagnostic accuracy, and although some studies suggest higher csPCa confirmation among MRI-positive units when image quality is better, that signal rests on very-low certainty evidence by the standards used to grade confidence in medical research findings. The author explicitly calls for prospective multicenter studies with unit-consistent reporting and complete two-by-two data, the minimal tabular structure needed to compute sensitivity, specificity, and their confidence intervals without ambiguity. The absence of a prospectively registered review protocol is acknowledged as a limitation, a transparency concession that, while common in the imaging literature, underscores the field’s broader methodological immaturity on this specific question.

The clinical stakes of this uncertainty are considerable. Multiparametric MRI has become the gatekeeper of the prostate cancer diagnostic pathway in many health systems, used both to decide which men should undergo biopsy and to guide targeted sampling of suspicious lesions. If image quality genuinely modulates cancer yield, then quality assurance programs, scanner upgrades, and standardized quality scoring could deliver measurable reductions in missed clinically significant cancers and unnecessary biopsies alike. Conversely, if the association is weak or confined to particular units of analysis, then expensive investments in imaging infrastructure may yield diminishing returns, and attention may need to shift toward interpretation, biopsy technique, or risk-based patient selection. The current evidence, as this review makes clear, cannot yet adjudicate between those futures.

What the review does establish, with unusual methodological candor, is how much work remains before the field can speak with one voice. Eight studies spanning heterogeneous populations, differing quality thresholds, varied reference standards, and incompatible units of analysis constitute a foundation too narrow for firm clinical guidance. The standardized outcome adopted here, the proportion of MRI-positive units with histopathologically confirmed csPCa, offers a template for future research that could finally allow meaningful synthesis across centers and countries. Until such studies arrive, the review’s message to radiologists, urologists, and policymakers is one of disciplined humility: the intuition that sharper images find more dangerous cancers is plausible and partially supported, but it is not yet proven, and the patients whose treatment depends on that assumption deserve evidence of higher certainty than the field has so far produced.

Subject of Research: The association between prostate MRI image quality and histopathologically confirmed clinically significant prostate cancer in MRI-positive examinations

Article Title: MRI quality and csPCa confirmation in MRI-positive examinations

Article References: Kesimal, U. (2026). MRI quality and csPCa confirmation in MRI-positive examinations. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02918-8

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02918-8

Keywords: prostate cancer, multiparametric MRI, PI-RADS, PI-QUAL, image quality, csPCa, systematic review, diagnostic accuracy, radiology, biopsy, QUADAS-2, BMC Medical Imaging

A new study published in BMC Cancer by Jian Shang, Yaolu Li, Lidan Yang, Donghui Zhang and colleagues at the Affiliated Cancer Hospital of Guangzhou Medical University set out to change that timeline. The research team developed and internally validated a machine-learning model that combines standard clinical variables with quantitative features extracted from multiparametric magnetic resonance imaging to predict LVI before a patient ever enters the operating room. The work, a retrospective single-center study, also explored whether the model’s output could stratify patients by their risk of disease-free survival, offering a glimpse of how preoperative imaging intelligence might one day guide gastric cancer management.

The technical foundation of the study is radiomics, a field that converts the pixel-and-voxel patterns inside medical images into hundreds or thousands of numerical descriptors. Where a radiologist’s eye might notice that a tumor looks heterogeneous or ill-defined, a radiomics pipeline quantifies exactly how heterogeneous, how textured, how intense, and how variable the tumor’s signal is across different imaging sequences. The researchers drew on three complementary MRI sequences: fat-suppressed T2-weighted imaging, which highlights tissue water content and edema; apparent diffusion coefficient maps derived from diffusion-weighted imaging, which reflect the restricted movement of water molecules in densely packed tumor tissue; and venous-phase contrast-enhanced MRI, which reveals how the tumor takes up gadolinium-based contrast agent through its blood supply.

Each sequence captures a different facet of tumor biology. Restricted diffusion, for example, often corresponds to high cellularity, a hallmark of aggressive malignancies. Contrast enhancement patterns can reflect the chaotic, leaky neovasculature that tumors build to feed themselves. T2 signal characteristics can distinguish mucin-rich or necrotic regions from solid cellular components. By mining all three sequences simultaneously, the team aimed to build a digital fingerprint of the tumor that correlates with the microscopic reality of vascular invasion.

The study population comprised 458 patients with gastric cancer, randomly assigned to a training cohort of 320 patients and an internal test cohort of 138. This split is a critical methodological safeguard: models built and evaluated on the same patients routinely overstate their own accuracy, so holding out a test set provides a more honest estimate of performance. From the MRI data, the researchers extracted radiomics features and applied feature reduction techniques, including least absolute shrinkage and selection operator regression, known as LASSO, which penalizes complexity and drives the coefficients of uninformative features to zero. After this filtering process, six radiomics features and four clinical predictors survived as the final model inputs.

The team then constructed three logistic-regression models to compare different predictor sets, designating the combined clinical-radiomics model as the primary model. Logistic regression may seem modest in an era of deep neural networks, but it offers transparency, calibration, and robustness with modest sample sizes, qualities that matter enormously in clinical prediction. Secondary comparisons using alternative machine-learning algorithms, including support vector machines and linear discriminant analysis, were performed with fold-restricted random oversampling, a technique that balances the classes within each cross-validation fold to avoid data leakage while addressing the inevitable imbalance between LVI-positive and LVI-negative patients.

The headline result: the combined clinical-radiomics model achieved an area under the receiver operating characteristic curve, or AUC, of 0.775 in the training cohort and 0.752 in the internal test cohort, with a bootstrap 95 percent confidence interval of 0.663 to 0.840. An AUC of 0.75 places the model in the territory of moderate discrimination, meaning it performs meaningfully better than chance but is far from infallible. At a sensitivity-oriented threshold of 0.217, chosen in the training set using the Youden index, the model detected LVI with a test sensitivity of 0.795, catching roughly four out of five invasion-positive cases, while its negative predictive value reached 0.855. In practical terms, when the model says a tumor is unlikely to have lymphovascular invasion, that reassurance is correct about 86 percent of the time.

Calibration, the question of whether predicted probabilities match observed reality, is often neglected in machine-learning studies but received careful attention here. In the test cohort, the calibration intercept was 0.092, the slope 0.965, and the Brier score 0.180, all indicating that the model’s probability estimates were reasonably honest rather than systematically overconfident or underconfident. Decision-curve analysis further assessed the clinical utility of the model across a range of threshold probabilities, evaluating the net benefit a patient population would gain if decisions were made according to the model’s predictions.

Perhaps the most intriguing finding is exploratory: the model’s output appeared to carry prognostic weight. Using an X-tile-derived cutoff of 0.59 established in the training data and applied unchanged to the test cohort, the researchers divided patients into risk groups for disease-free survival. In an adjusted complete-case Cox proportional hazards model, each standard deviation increase in the model’s predicted probability was associated with a hazard ratio of 1.57 for worse disease-free survival, with a 95 percent confidence interval of 1.21 to 2.03. A single preoperative number, computed from scans and blood markers, appeared to encode information about how the disease would unfold, independent of the adjustment variables included in the analysis.

The authors are appropriately measured about what their model can and cannot do. They emphasize that the model is not ready for stand-alone clinical use and requires prospective external validation in independent centers before any deployment. The study was retrospective and single-center, meaning the MRI scanners, imaging protocols, and patient demographics of one institution shaped the data. Radiomics features are notoriously sensitive to variations in acquisition parameters, and a model trained in Guangzhou may not transfer cleanly to a hospital with different equipment. The disease-free survival findings, too, are explicitly labeled exploratory, and the sensitivity analyses using a training-median cutoff underscore the uncertainty around how the risk stratification should be operationalized.

Still, the study adds to a growing body of evidence that the information needed to characterize a tumor’s aggressiveness may already be hiding inside routine imaging. Every gastric cancer patient undergoing staging typically receives cross-sectional imaging; the marginal cost of running a radiomics algorithm on those existing scans is nearly zero. If future prospective studies confirm these results, a high negative predictive value at a sensitivity-oriented threshold suggests a plausible triage role: patients whose scans argue strongly against LVI might be candidates for less aggressive perioperative planning, while those flagged as high risk could be prioritized for intensified neoadjuvant therapy, extended lymph node dissection, or closer surveillance. The vision is not to replace pathologists but to move critical biological knowledge upstream in the treatment timeline, from the postoperative pathology report to the preoperative clinic visit.

The work also highlights a broader trend in oncology: the convergence of interpretable machine learning, multiparametric imaging, and clinical data into decision-support tools that are auditable rather than opaque. By retaining logistic regression as the primary model and reporting calibration alongside discrimination, the Guangzhou team modeled the kind of methodological discipline that regulatory science increasingly demands. The path from a 0.75 AUC in a single center to a validated clinical tool is long, but this study maps an early stretch of it, showing that the vessels a tumor invades in secret may leave traces visible to an algorithm long before any scalpel is lifted.

Subject of Research: Preoperative prediction of lymphovascular invasion in gastric cancer using multiparametric MRI radiomics and machine learning

Article Title: Preoperative prediction of lymphovascular invasion and prognostic risk stratification in gastric cancer using multiparametric MRI radiomics and machine learning: a retrospective model development and internal validation study

Article References: Preoperative prediction of lymphovascular invasion and prognostic risk stratification in gastric cancer using multiparametric MRI radiomics and machine learning: a retrospective model development and internal validation study. (n.d.). https://doi.org/10.1186/s12885-026-17042-7

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17042-7

Keywords: gastric cancer, lymphovascular invasion, multiparametric MRI, radiomics, machine learning, logistic regression, disease-free survival, predictive modeling, medical imaging, cancer prognosis, internal validation, LASSO

The core problem lies in what clinicians call the diagnostic gray zone. Serum prostate-specific antigen (PSA), the workhorse of prostate cancer screening since the early 1990s, loses much of its discriminative power when values fall between 4 and 10 ng/mL, a range where benign prostatic hyperplasia and prostatitis routinely elevate the marker just as aggressively as tumors do. Multiparametric MRI, the other pillar of modern triage, has its own blind spot: lesions graded PI-RADS category 3, the equivocal middle of the Prostate Imaging Reporting and Data System, carry genuinely ambiguous cancer probability. In this gray zone, the overlap between malignant and benign disease drives a substantial number of unnecessary biopsies—procedures that carry real morbidity, from bleeding and infection to the psychological burden of waiting, and that frequently detect indolent cancers a patient might never have needed to know about.

The review’s authors, led by Zijie Nie and corresponding author Lingling Wu, organize the emerging solutions into three broad categories: liquid-based molecular biomarkers, advanced imaging modalities, and integrative frameworks that fuse multiple data streams. On the biomarker front, the most mature tools refine what PSA already tells us. The Prostate Health Index (PHI), a mathematical combination of total PSA, free PSA, and the precursor form p2PSA, has been validated across European and Asian populations, with recent large Chinese cohort data confirming its utility and newer derivatives such as PHI density—PHI divided by prostate volume—outperforming PSA density in equivocal MRI settings. The 4Kscore, a four-kallikrein blood panel, has shown its worth as a reflex test after elevated PSA; in the GÖTEBORG-2 screening trial it performed well as a second-line discriminator, and the ProScreen randomized trial demonstrated that combining PSA, the kallikrein panel, and MRI can streamline screening pathways.

Perhaps the most conceptually elegant of the refined PSA tools is IsoPSA, which abandons concentration measurement altogether. Instead of asking how much PSA is in the blood, IsoPSA interrogates the structural heterogeneity of the PSA protein itself—cancerous tissue produces subtly different molecular forms of the enzyme than benign tissue does. Prospective multicenter validation showed improved detection of both any cancer and high-grade disease compared with conventional total PSA, and notably, its performance characteristics are unaffected by 5-alpha reductase inhibitors and alpha-blockers, drugs that confound standard PSA interpretation. A recent study further reported that IsoPSA density improves risk stratification and biopsy decision-making for clinically significant cancer.

Urine-based tests exploit a different biological logic: tumor-derived molecular signatures can be captured non-invasively from prostatic fluid expressed during a digital rectal examination or simply from first-catch urine. SelectMDx, a two-gene mRNA test, has been validated in prospective multicenter studies of biopsy-naïve men and compared directly against mpMRI in diagnostic meta-analyses. MyProstateScore (MPS), built around the TMPRSS2:ERG fusion gene and PCA3, has demonstrated a robust ability to rule out clinically significant cancer, including in men with equivocal PI-RADS 3 lesions—precisely the population where the clinical need is greatest. Its successor, MyProstateScore 2.0, an 18-gene urine test validated in JAMA Oncology, extends this approach to high-grade cancer detection using first-catch urine without any prior examination, removing a logistical barrier to widespread use.

Exosome technology represents a further step along the liquid biopsy frontier. The ExoDx Prostate IntelliScore test analyzes gene expression in urinary exosomes—nanoscale vesicles shed by prostate tissue—to predict high-grade cancer at initial biopsy in men with PSA between 2 and 10 ng/mL, with clinical performance confirmed across three independent prospective studies. Beyond RNA, researchers are mining exosomes for proteins and metabolites: proteomic profiling of urinary large extracellular vesicles, chemical affinity capture of plasma extracellular vesicles for large-scale biomarker discovery, and metabolomic fingerprints of vesicles from prostatic fluid all feature in the review as rapidly maturing avenues. Emerging candidates such as annexin A3, the homeobox protein engrailed-2 (EN2), CRISP3, and the cholinergic peptide SLURP1 add to a growing catalog of tumor-specific molecules detectable in urine.

MicroRNAs and metabolomics round out the molecular picture. Circulating and urinary microRNAs—including miR-21, miR-145, miR-30b-3p, and miR-375—have shown the ability to distinguish prostate cancer from benign lesions, with novel detection platforms such as silver nanoparticle sensors pushing toward point-of-care measurement. Metabolomics takes aim at the tumor’s altered biochemistry: sarcosine in urine, lipid species in exosomes, and serum metabolite panels have all discriminated cancer from benign prostatic hyperplasia within the PSA gray zone, and a recent systematic review of pre-diagnostic untargeted metabolomics supports their prospective value. The biological rationale is sound, since prostate cancer undergoes profound metabolic reprogramming, and metabolites often change earlier than structural imaging can detect.

On the imaging side, the review catalogues technologies that go beyond conventional mpMRI. Biparametric and abbreviated MRI protocols promise wider access with comparable detection accuracy. Magnetic resonance elastography measures the physical stiffness of tissue—cancers are typically harder than benign tissue—and tomoelastography based on multifrequency MRE has outperformed mpMRI in head-to-head comparisons. Amide proton transfer-weighted imaging, a chemical exchange saturation transfer technique, probes tissue pH and protein content at the molecular level and has added value to PI-RADS v2.1 in detecting clinically significant disease. Micro-ultrasound, with resolution roughly three times finer than standard ultrasound, performed comparably to MRI in biopsy-naïve men and was tested head-to-head against MRI-guided biopsy in the OPTIMUM randomized trial. Most striking is PSMA PET/CT, which images the prostate-specific membrane antigen expressed on tumor cells; studies such as PRIMARY have shown its additive value to mpMRI triage, and [18F]DCFPyL PET/CT has reduced unnecessary biopsies in PI-RADS 3/4 patients—though benign prostatic hyperplasia-related false positives remain its acknowledged Achilles’ heel.

The review’s central thesis, however, is that no single test will win. Instead, the field is converging on multimodal integration. Risk calculators such as Stockholm3—which combines protein biomarkers, genetic variants, and clinical variables—have been validated in multiethnic cohorts and combined productively with MRI. PSA density and PHI density refine biopsy thresholds when layered onto PI-RADS scores, and combining the PRIMARY score with PSA density helps avoid unnecessary biopsies after negative mpMRI. Artificial intelligence is the connective tissue of this shift: deep learning models detect clinically significant cancer on MRI with performance approaching that of expert radiologists, as demonstrated in the international PI-CAI study published in Lancet Oncology, while multimodal AI systems that fuse clinical data, biomarkers, and imaging inputs have improved detection beyond any single modality. Radiomics—high-throughput extraction of quantitative image features—adds another layer, with machine learning models predicting malignancy even in equivocal PI-RADS 3 lesions.

The implications for patients are concrete. By sharpening the boundary between cancer and its benign mimics before biopsy, these tools could spare thousands of men annually from invasive procedures that carry morbidity and psychological cost, while directing biopsies more precisely toward those with clinically significant disease. The authors emphasize that this is a paradigm shift from single-parameter screening to multimodal, AI-assisted decision-making—personalized risk stratification built from blood, urine, and image data. Challenges remain, including validation across diverse populations, cost, and clinical implementation, but the trajectory is clear: the diagnostic gray zone that has frustrated urologists for three decades is finally being mapped, molecule by molecule and pixel by pixel, and the biopsy needle may soon arrive only where it is truly needed.

Subject of Research: Non-invasive pre-biopsy diagnostic strategies for discriminating prostate cancer from benign prostatic diseases

Article Title: Recent advances in discriminating prostate cancer from benign prostatic diseases in the pre-biopsy population

Article References: Nie, Z., Zhu, T., Chen, Y., Zhang, W., & Wu, L. (2026). Recent advances in discriminating prostate cancer from benign prostatic diseases in the pre-biopsy population. Medical Oncology, 43(11), Article 305. https://doi.org/10.1007/s12032-026-03424-1

Image Credits: AI Generated

DOI: 10.1007/s12032-026-03424-1

Keywords: prostate cancer, PSA, liquid biopsy, multiparametric MRI, biomarkers, PI-RADS, artificial intelligence, PSMA PET, urine tests, risk stratification, prostate biopsy, metabolomics

The clinical dilemma the study addresses is well known to urologists and nuclear medicine physicians alike. Multiparametric MRI has become the cornerstone of modern prostate cancer detection, allowing clinicians to localize suspicious lesions, direct biopsy needles toward the most concerning regions, and spare many men unnecessary sampling of slow-growing tumors that may never require treatment. Yet the technology is not infallible. According to the researchers, up to 20 to 30 percent of clinically significant prostate cancers may be missed in patients whose MRI scans are negative or equivocal. For these men, a reassuring scan does not actually rule out dangerous disease, and the standard response—repeated or extended systematic biopsies—carries its own burden of bleeding, infection, anxiety, and cost, all while aggressive tumors can still slip through the sampling grid.

“A negative or unclear MRI does not completely rule out clinically significant prostate cancer. As a result, some men may undergo repeated or extensive biopsies, while relevant tumors may still be missed,” said Andrea Farolfi, MD, a nuclear medicine physician at the Nuclear Medicine Unit of IRCCS Azienda Ospedaliero-Universitaria di Bologna and the study’s lead author. “Determining whether PSMA PET can refine risk stratification and guide targeted sampling for these men could have a direct impact on clinical decision-making.” The rationale behind the approach rests on the biology of the prostate-specific membrane antigen, or PSMA, a protein that is overexpressed on the surface of prostate cancer cells, and often to a far greater degree in aggressive, high-grade disease than in indolent tissue.

PSMA PET/CT exploits this molecular signature by pairing a radiolabeled ligand that binds to PSMA with the anatomic precision of computed tomography. When the radiotracer accumulates intensely within a prostatic region, it signals not merely anatomic distortion, as MRI might show, but active metabolic expression of a marker tightly linked to tumor aggressiveness. In the FUPERMAN study, the researchers enrolled 63 men with negative or inconclusive multiparametric MRI scans who nonetheless carried persistent clinical suspicion of prostate cancer. Each participant underwent gallium-68 PSMA-11 PET/CT followed by prostate biopsy performed with PET fusion guidance, meaning the targeting information from the molecular image was directly co-registered with the biopsy system in real time. A subset of 11 patients additionally underwent dynamic total-body PET, an emerging technique that captures the kinetic behavior of the tracer as it flows into and binds within tissue.

The results were striking in their consistency. PSMA PET/CT was positive in 25 of the 26 men who were ultimately diagnosed with clinically significant prostate cancer, indicating that the molecular image rarely overlooked disease that biopsy later confirmed. PSMA PET/CT-targeted biopsy alone detected clinically significant cancer in 37 percent of the entire study population, while systematic biopsy alone managed only 21 percent. Combining the two strategies raised the detection rate to 41 percent, demonstrating that the molecularly guided approach does not simply duplicate what systematic sampling finds—it uncovers cancers that the conventional grid-based method misses, particularly in prostatic zones where MRI provided no suspicious target to aim at.

Beyond simple detection, the study explored whether quantitative information from the PET scan itself could predict which patients harbored aggressive disease. The researchers found that SUVmax, the maximum standardized uptake value reflecting the peak concentration of radiotracer in a lesion, was independently associated with clinically significant prostate cancer. In other words, the intensity of the molecular signal carried prognostic weight, offering a semiquantitative bridge between imaging and pathology that could help clinicians decide which men truly need biopsy and which might be safely monitored. This kind of risk stratification is precisely what the field has been seeking for the MRI-negative population, where current guidelines offer limited guidance and clinical judgment often defaults to repeat invasive procedures.

The exploratory dynamic total-body PET component added a further layer of technical sophistication. By acquiring continuous imaging data immediately after tracer injection, dynamic total-body PET allows researchers to characterize the kinetics of radiotracer uptake and washout within prostatic lesions. In the 11 patients studied with this protocol, the investigators observed distinct kinetic patterns in aggressive lesions, particularly in cases where MRI findings were unclear. While the subset was small and the analysis explicitly exploratory, the finding hints at a future in which the temporal behavior of PSMA binding—how quickly tracer accumulates and how persistently it is retained—could complement static uptake measurements and further sharpen the distinction between indolent and clinically threatening disease.

“Our findings suggest that PSMA PET/CT could provide an additional tool for identifying men who remain at risk despite non-diagnostic MRI findings and for directing the biopsy toward the most suspicious area,” Farolfi said. “If confirmed in larger studies, this approach could make the diagnostic pathway more accurate and personalized.” The implications for patient care extend beyond the immediate detection statistics. For men caught in the diagnostic gray zone—elevated PSA, negative or equivocal MRI, and mounting pressure for repeat biopsy—a positive PSMA PET/CT could concentrate sampling where it matters most, potentially reducing the number of biopsy cores required, lowering complication rates, and sparing men with truly negative molecular imaging from yet another round of invasive sampling.

The study, titled “PSMA PET/CT–Targeted Biopsy in Men with Negative or Equivocal Multiparametric MRI and Exploratory Dynamic Total-Body PET: The FUPERMAN Study,” was conducted by a multidisciplinary team spanning nuclear medicine, urology, radiology, and internal medicine at the University of Bologna and its teaching hospital. The collaboration reflects the practical demands of the technique itself, which requires nuclear medicine physicians to interpret molecular images, radiologists to reconcile them with prior MRI data, and urologists to execute fusion-guided biopsies with millimeter-level precision. The research was published in The Journal of Nuclear Medicine, the flagship journal of the Society of Nuclear Medicine and Molecular Imaging, and carries the DOI 10.2967/jnumed.126.272108.

As with any single-center study of modest size, the findings will need validation in larger, multi-institutional cohorts before PSMA PET/CT-guided biopsy becomes a standard recommendation for the MRI-negative population. Questions remain about cost-effectiveness, radiation exposure, access to PSMA radiotracers, and whether the impressive detection rates hold across more diverse patient populations. Nevertheless, the FUPERMAN results mark an important proof of concept: a molecular imaging modality originally developed for staging advanced disease can, when fused with biopsy guidance, illuminate cancers that the anatomical gold standard cannot see. For the substantial minority of men whose MRI scans fail them, that could mean the difference between an aggressive tumor caught early and one discovered only after it has progressed beyond the gland.

Subject of Research: PSMA PET/CT-guided targeted biopsy for detecting clinically significant prostate cancer in men with negative or equivocal MRI

Article Title: PSMA PET/CT-targeted biopsy improves detection of clinically significant prostate cancer

Article References: PSMA PET/CT-targeted biopsy improves detection of clinically significant prostate cancer. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: PSMA PET/CT, prostate cancer, targeted biopsy, multiparametric MRI, clinically significant cancer, total-body PET, SUVmax, risk stratification, nuclear medicine, FUPERMAN study, fusion-guided biopsy, diagnostic imaging

A new study published in the European Journal of Nuclear Medicine and Molecular Imaging suggests that a hybrid imaging technique may finally bring quantitative rigor to this assessment. Researchers at Peking Union Medical College Hospital in Beijing prospectively enrolled 33 patients with thyroid-associated ophthalmopathy and scanned them on an integrated PET/MRI system using a radiotracer called 18F-NOTA-FAPI-04. Unlike conventional FDG, which tracks glucose metabolism in any active cell, FAPI tracers bind to fibroblast activation protein, a molecule displayed by activated fibroblasts and their inflammatory partners. Because thyroid eye disease is fundamentally a fibroinflammatory disorder of the orbital tissues, the tracer offers a way to visualize the cellular machinery driving the disease rather than its downstream anatomical consequences.

The technical logic of the study is worth unpacking. Each patient’s orbits were imaged simultaneously with PET, which quantifies molecular uptake, and with a battery of advanced MRI sequences that probe tissue microstructure. The MRI panel included T2 mapping, which reflects tissue water content and therefore edema; intravoxel incoherent motion diffusion imaging, which separates true water diffusion from blood flow within capillaries; and diffusion kurtosis imaging, which captures deviations from simple Gaussian diffusion and thus hints at tissue complexity. From these sequences the team extracted parametric maps of normalized T2 signal, the heterogeneity index alpha, the distributed diffusion coefficient, the diffusion coefficient, relative blood flow, mean diffusivity, and mean kurtosis. The goal was to see whether the molecular signal from PET and the microstructural signal from MRI tell the same story, and whether combining them outperforms either alone.

Participants were classified using the Clinical Activity Score, with eyes scoring three or higher considered active and those below three considered inactive. At the patient level, a person was labeled inactive only if both eyes were inactive. The researchers then extracted a rich set of quantitative features from each eye: PET metrics including maximum and mean standardized uptake values, metabolic tumor volume, and total lesion FAPI uptake, plus first-order histogram statistics such as mean, skewness, and kurtosis from each MRI parametric map. A support vector machine, a standard machine-learning classifier, was trained to distinguish active from inactive disease, and performance was evaluated with receiver operating characteristic analysis.

Reliability came first, and the numbers were reassuring. Interobserver and intraobserver agreement in delineating the orbital regions of interest was strong, with Dice similarity coefficients of 0.850, 0.880, and 0.901 for the different comparisons. Every quantitative PET and MRI parameter showed excellent reproducibility, with intraclass correlation coefficients above 0.907. In a field where subjective eyeballing of scans has long been the norm, this level of measurement stability matters: it means the features being fed into the classifier are not artifacts of who happened to draw the contours.

The correlation analysis produced one of the study’s most interesting nuances. At the patient level, where both eyes are averaged together, PET and MRI features correlated only weakly. But when each eye was analyzed as its own unit, the correlations strengthened considerably, particularly between PET uptake metrics and features derived from T2 mapping, the heterogeneity index, the diffusion coefficient, and mean diffusivity. This asymmetry makes biological sense. Thyroid eye disease is notoriously asymmetric, with one eye often far more inflamed than the other, and averaging across eyes dilutes the very signals that matter. The finding is a quiet argument for eye-level, rather than patient-level, imaging assessment in orbital disease.

The group comparisons confirmed that several quantitative features separate active from inactive disease. At the eye level, mean standardized uptake value, metabolic tumor volume, and total lesion FAPI uptake were all significantly higher in active eyes, while maximum standardized uptake value alone showed no significant difference, a reminder that peak values are often less informative than volume-weighted averages. On the MRI side, features from T2 mapping, the heterogeneity index, the diffusion coefficient, mean diffusivity, and mean kurtosis also differed significantly between active and inactive groups, consistent with the idea that active disease carries more edema and altered diffusion characteristics than fibrotic, quiescent tissue.

When it came to discriminating activity, the combined approach won. Among single PET parameters, mean standardized uptake value performed best at the patient level with an area under the curve of 0.759, while metabolic tumor volume led at the eye level with an AUC of 0.767. These are respectable but unremarkable figures, in the range of many clinical biomarkers. The combined PET/MRI model, however, pushed performance to an AUC of 0.844 at the patient level and 0.859 at the eye level. The improvement over either modality alone indicates that FAPI uptake and MRI-derived microstructural features carry complementary, not redundant, information about the state of the orbital tissues. In practical terms, the PET signal appears to capture the cellular fibroinflammatory process while the MRI parameters capture its tissue-level consequences, and the classifier benefits from having both.

The clinical implications are significant. Treatment for active thyroid eye disease, including high-dose glucocorticoids and newer targeted agents such as anti-IGF-1 receptor antibodies, works best early in the inflammatory phase and offers little once fibrosis has set in. Conversely, rehabilitative surgeries are best deferred until the disease is inactive. A quantitative imaging biomarker that reliably separates these phases could spare patients from ineffective treatment, reduce exposure to steroid side effects, and help time interventions more precisely. It could also serve as an objective endpoint in clinical trials, where the Clinical Activity Score’s known inter-observer variability has long complicated the interpretation of results. Prior studies have explored FDG-PET and multiparametric MRI separately for this purpose, but the FAPI tracer’s specificity for activated fibroblasts, combined with simultaneous MRI acquisition on a single scanner, represents a meaningful step forward in what the field calls molecular-radiological phenotyping.

Cautions remain, and the authors are candid about them. Thirty-three patients is a small cohort, and the classifier’s performance will need validation in larger, independent, and ideally multi-center populations before it can influence routine care. The Clinical Activity Score itself, used here as the reference standard, is an imperfect ground truth, which means the imaging model is being trained to reproduce a clinical judgment rather than to measure disease biology directly. The study was registered as a clinical trial and conducted under ethics approval at Peking Union Medical College Hospital, and the team notes that further validation is required. Still, the trajectory is clear: a single integrated scan that fuses molecular information about fibroblast activity with quantitative maps of tissue edema, perfusion, and diffusion could transform thyroid eye disease from a condition assessed by counting symptoms into one measured, monitored, and treated on the basis of what is actually happening inside the orbit. For a disease that can quietly steal sight, that kind of clarity cannot come soon enough.

Subject of Research: Quantitative assessment of disease activity in thyroid-associated ophthalmopathy using integrated 18F-NOTA-FAPI-04 PET/MRI

Article Title: Integrated 18F-NOTA-FAPI-04 PET/MRI for quantitative assessment of disease activity in thyroid-associated ophthalmopathy

Article References: Yang, X., Gan, L., Shi, X., Wu, M., Li, E., Zhang, Y., Hao, Z., Huang, Z., Xing, H., Liu, X., & Huo, L. (2026). Integrated 18F-NOTA-FAPI-04 PET/MRI for quantitative assessment of disease activity in thyroid-associated ophthalmopathy. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08157-x

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08157-x

Keywords: thyroid-associated ophthalmopathy, thyroid eye disease, PET/MRI, 18F-NOTA-FAPI-04, fibroblast activation protein, Clinical Activity Score, multiparametric MRI, intravoxel incoherent motion, diffusion kurtosis imaging, radiomics, support vector machine, Graves orbitopathy

The study, led by researchers at the First Affiliated Hospital of Soochow University in Suzhou, China, was designed as a paired diagnostic accuracy trial registered with the Chinese Clinical Trial Registry. Every participant underwent all three biopsy strategies in a single session: a tri-modal PSMA PET/CT-MRI-US fusion biopsy, a conventional MRI-US fusion biopsy, and a 12-core systematic biopsy. Because each man served as his own control, the comparison directly quantified the incremental yield contributed by adding molecular imaging to the targeting pipeline, rather than relying on comparisons between different groups of patients. The primary endpoint was the absolute difference in patient-level detection of clinically significant prostate cancer, defined as Grade Group 2 or higher on the International Society of Urological Pathology scale.

The biological rationale for the approach is elegant. Multiparametric MRI detects structural disruption—changes in tissue architecture, water diffusion, and contrast enhancement—whereas PSMA PET targets the overexpression of prostate-specific membrane antigen, a receptor that is upregulated on aggressive tumor cells. The two modalities therefore interrogate fundamentally different features of the same disease, and published discordance rates between them range from 30 to 50 percent. Lesions visible only on PSMA PET carry clinically significant cancer rates of roughly 26 to 39 percent, while MRI-only lesions carry rates of 15 to 20 percent. In this cohort, the asymmetry was even starker: among non-co-localized lesions, 39.3 percent of PSMA-only lesions harbored clinically significant cancer, compared with just 17.5 percent of MRI-only lesions.

Translating those molecular signals into accurate needle placement, however, demands precise spatial registration between preoperative images and the real-time ultrasound plane. Conventional MRI-US fusion relies on software-assisted alignment of images acquired in different positions—the patient supine during MRI and in the lithotomy position during transrectal ultrasound—and registration errors from prostate deformation, patient movement, and operator variability can exceed the diameter of small aggressive tumors. The tri-modal platform, developed under a Chinese patent and implemented on the Carbon AI fusion system, attacks this problem in three stages. A convolutional neural network first segments the prostate on preoperative CT and MRI, converting the resulting surfaces into point clouds that are rigidly aligned using an iterative closest point algorithm. The optimal transformation minimizes a combined loss function balancing mutual information between the images and Dice overlap between the segmented surfaces.

Because PSMA PET and CT share the same coordinate system, the PET signal can then be mapped directly into MRI space. During the biopsy itself, the same neural network segments the real-time transrectal ultrasound images, and point-cloud registration under electromagnetic tracking solves for the rotation and translation that best match the live prostate contour to its MRI counterpart. The result is real-time, three-modality navigation in which molecular uptake is converted directly into biopsy coordinates. Three cores were taken from each PET-positive lesion and three from each MRI-positive lesion, meaning co-localized lesions received six cores, while systematic biopsy followed the standard 12-core Ginsburg protocol. Pathologists reading the cores were blinded to all imaging data and biopsy approach.

The headline result was unambiguous. Tri-modal fusion biopsy found clinically significant cancer in 119 of the 308 patients, versus 104 detected by MRI-US fusion biopsy. Fifteen men were identified exclusively by the tri-modal approach, and not a single cancer was found by MRI-US fusion alone that the tri-modal method missed. The number needed to biopsy to detect one additional clinically significant cancer was 20.5. Subgroup analyses revealed where the extra yield came from: PSMA-positive lesions smaller than 10 millimeters showed an 8.3 percent absolute gain, PI-RADS 4 lesions a 5.8 percent gain, and patients with prostate-specific antigen density of 0.15 or higher a 6.0 percent gain. For PI-RADS 5 lesions, the advantage disappeared—a ceiling effect, since MRI-guided biopsy already detects nearly all cancers in that category.

Lesion-level statistics reinforced the message. Using generalized estimating equations to account for multiple lesions within each patient, PSMA-positive lesions carried nearly twice the odds of harboring clinically significant cancer compared with MRI-positive lesions, an association that persisted after adjustment for age, PSA density, and PI-RADS score, reaching an odds ratio of 1.97. Co-localized lesions—positive on both modalities—had the highest cancer rate of all at 62.3 percent. Notably, the modality-by-zone interaction was not significant, indicating that the heterogeneous signal typical of the transition zone, where benign enlargement can mimic disease, did not compromise the fusion accuracy. Among the 124 men with clinically significant cancer, only five were detected solely by systematic biopsy, invisible on both PET and MRI.

Beyond the biopsy table, the team built a PSMA-derived risk score combining lesion diameter, maximum standardized uptake value, the SUV ratio relative to the parotid gland, lesion location, and diffuse uptake patterns. In a held-out test set of 92 patients, the score achieved an area under the curve of 0.933, outperforming both the visual PRIMARY score (0.873) and the molecular imaging miPSMA score (0.838), with good calibration confirmed by the Hosmer-Lemeshow test. When the researchers combined the score with PI-RADS category and PSA density into a clinical model, a simulated risk-guided strategy would have spared biopsies in roughly 42 percent of the test-set patients while missing only 2.3 percent of clinically significant cancers—below the 5 percent miss threshold recommended by current guidelines—and producing no overdiagnoses.

The authors are careful to frame that simulation as exploratory. The risk thresholds are cohort-specific, the test set was small, and the analysis was retrospective rather than prospective. Other limitations deserve emphasis. The superset design meant the tri-modal approach sampled all MRI-visible lesions plus additional PSMA-only targets, so the 4.9 percent detection gain reflects both molecular targeting and increased sampling density—321 extra cores in total. The study also cannot disentangle how much of the benefit came from the AI registration platform versus the PET information itself, and needle-guidance accuracy was not formally validated. As a single-center study, the results await multicenter confirmation before any change to clinical practice.

Safety and cost considerations round out the picture. No serious adverse events occurred; transient hematuria affected a third of patients on day one but fell to about 1 percent within a week, and only two men developed brief post-procedural fever. Adding PSMA PET/CT does carry roughly 3 to 5 millisieverts of radiation—comparable to a year of natural background—and meaningful cost, which is why the modality is currently reserved for equivocal MRI findings or high-risk disease rather than routine pre-biopsy use. Yet if the simulated risk-guided pathway holds up in prospective validation, the scans spared in the 42 percent of low-risk men could offset much of that burden. For now, the study offers a technically sophisticated preview of where prostate cancer diagnosis may be heading: a fusion of anatomy, function, and molecular biology, guided by algorithms, that finds the tumors MRI alone cannot see.

Subject of Research: Tri-modal PSMA PET/CT-MRI-US fusion biopsy for detecting clinically significant prostate cancer

Article Title: Tri-modal PSMA PET/CT-MRI-US fusion biopsy for clinically significant prostate cancer: a prospective paired diagnostic study

Article References: Huang, C., Jin, L., Sun, Y., Wu, M., Zhang, X., Xu, X., He, X., Qiu, F., Wang, X., Guo, L., Huang, R., Li, J., Deng, S., Lin, Y., Wei, X., Zhang, B., & Huang, Y. (2026). Tri-modal PSMA PET/CT-MRI-US fusion biopsy for clinically significant prostate cancer: a prospective paired diagnostic study. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08181-x

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08181-x

Keywords: prostate cancer, PSMA PET/CT, fusion biopsy, multiparametric MRI, diagnostic accuracy, AI registration, electromagnetic tracking, risk score, PI-RADS, systematic biopsy, molecular imaging, clinical trial

The study, published in the European Journal of Nuclear Medicine and Molecular Imaging, enrolled fifty men with biopsy-proven high-risk prostate cancer scheduled for robot-assisted radical prostatectomy between January 2022 and July 2024. Every participant underwent both [18F]F-PSMA-1007 PET/CT and [68Ga]Ga-PSMA-11 PET/CT before surgery, and forty-five also received multiparametric MRI after a protocol amendment added it as an exploratory third modality. The researchers then mapped tumours across six anatomical sextants of the prostate—base, mid-gland and apex on each side—using whole-mount histopathology of the entirely embedded surgical specimen, in which the gland is sectioned into 5 to 6 millimetre slices, embedded in paraffin, cut at 4 to 5 micrometres and stained for microscopic examination by two blinded uropathologists.

The underlying biology makes PSMA PET a compelling candidate for this task. Prostate-specific membrane antigen is a protein abundantly displayed on the surface of prostate cancer cells, and radiolabelled ligands that bind it allow clinicians to visualise tumour deposits as bright focal spots against dimmer background tissue. The two tracers differ chemically: gallium-68 is produced in a generator and has a half-life of just over an hour, while fluorine-18 offers a longer half-life of about 110 minutes, lower positron energy that yields sharper images, and reduced urinary excretion that can keep the bladder from obscuring the prostate. These theoretical advantages have fuelled a long-standing debate about which ligand clinicians should prefer.

The Danish team’s answer, at least within this cohort, is that it barely matters. At the patient level, [18F]F-PSMA-1007 PET/CT detected clinically significant prostate cancer—defined as grade group 3 or higher, or grade group 2 with a tertiary Gleason pattern 5—with 98 percent sensitivity and 96 percent diagnostic accuracy. [68Ga]Ga-PSMA-11 achieved 100 percent sensitivity and 98 percent accuracy, while MRI reached 96 percent sensitivity and 93 percent accuracy. No statistically significant difference separated the two PSMA ligands. The single apparent false-positive PET finding turned out to reflect tracer uptake in lower-grade, clinically insignificant cancer rather than benign tissue, underscoring how faithfully PSMA signal tracks with malignant biology even when it overshoots the clinical threshold.

The regional analysis told a more nuanced story. Of 300 prostate sextants, 205 contained clinically significant cancer. [18F]F-PSMA-1007 correctly identified clinically significant disease in 143 sextants, yielding a mixed-model accuracy of 69.8 percent, while [68Ga]Ga-PSMA-11 detected 136 sextants with 67.8 percent accuracy. MRI, evaluated across 270 sextants, achieved 62.9 percent accuracy. The statistical modelling, which used mixed-effects regression with a random intercept per patient to account for the clustering of repeated measurements within each prostate, revealed a striking trade-off: both PSMA tracers were significantly more sensitive than MRI, at roughly 70 percent versus 51 percent, but MRI was significantly more specific, at nearly 89 percent versus about 68 percent for the PET ligands. In other words, PSMA PET finds more of the tumour, while MRI makes fewer false alarms.

That complementarity is the study’s most clinically resonant finding. MRI, the current guideline-endorsed standard for tumour localisation and surgical planning, excels at anatomical detail but can miss lesions obscured by motion artefacts, hip prostheses, or inflammation, and it systematically underestimates tumour extent. PSMA PET, by contrast, lights up malignant tissue with high tumour-to-background contrast but flags some regions that turn out to harbour only low-grade disease. The authors argue that the two technologies may work best together, a conclusion that aligns with emerging multimodal approaches in which artificial intelligence models integrate PSMA PET, MRI, and clinical variables to predict adverse pathology before surgery.

Where both technologies stumbled was in staging tumour extension beyond the prostate gland. Extraprostatic extension was present in 26 of the 50 patients, yet sensitivity for detecting it was a dismal 12 percent for [18F]F-PSMA-1007, 23 percent for [68Ga]Ga-PSMA-11, and 9 percent for MRI. Seminal vesicle invasion, present in ten patients, fared somewhat better but still poorly, with both PET tracers reaching 30 percent sensitivity and MRI 14 percent—though specificity was nearly perfect, meaning that when any modality did call invasion, it was almost always right. These sobering numbers suggest that neither PSMA PET nor MRI is ready to replace histopathological assessment for local T-staging, and that surgical decisions about nerve-sparing must still rest on a synthesis of imaging, biopsy data and clinical judgement.

The study’s methodological rigour deserves emphasis. It followed STARD reporting guidelines, was registered in the EudraCT database, and was monitored under Good Clinical Practice. Readers were blinded to clinical data, histopathology and the other imaging modalities, and each reader interpreted only a single modality, with equivocal findings resolved by consensus. The researchers applied the PROMISE standardised reporting criteria for PSMA PET, using the liver as the reference background for the gallium tracer and the spleen for the fluorine tracer, and scored MRI lesions with PI-RADS version 2.1, counting regions as positive only at scores of 4 or higher. A neighbouring-region approach was used to minimise spatial mismatch between in-vivo imaging and ex-vivo pathology, acknowledging the tissue deformation and shrinkage that inevitably occur between scanning and sectioning.

The authors are careful about what their results do and do not prove. The study was designed to detect differences between modalities, not to formally demonstrate equivalence, so the absence of a statistically significant gap between the two PSMA ligands should not be read as proof of interchangeability. The single-centre design, the modest sample size, and the enrolment of patients from a high-volume tertiary centre all limit generalisability, and the MRI comparison was exploratory rather than prespecified. Still, the head-to-head design within the same patients, validated against whole-mount histopathology, is rare, and the findings provide hypothesis-generating evidence that both ligands perform comparably for intraprostatic localisation.

The implications ripple outward as PSMA PET expands from its established role in staging recurrent and metastatic disease toward primary tumour characterisation. If fluorine-18 and gallium-68 tracers deliver comparable intraprostatic performance, supply chains, costs and local production capacity may weigh more heavily in tracer selection than diagnostic nuance. Meanwhile, the demonstrated sensitivity advantage of PSMA PET over MRI, paired with MRI’s specificity edge, strengthens the case for hybrid and multimodal strategies that fuse molecular and anatomical information. As deep-learning models increasingly promise individualised risk stratification from combined imaging, this study supplies an essential foundation: a carefully validated account of what each modality, and each tracer, can and cannot see inside the prostate gland.

Subject of Research: Head-to-head comparison of two PSMA PET tracers and MRI for detecting and localising clinically significant prostate cancer against histopathology

Article Title: Head-to-head comparison of [¹⁸F]F-PSMA-1007 PET/CT and [⁶⁸Ga]Ga-PSMA-11 PET/CT for intraprostatic tumour detection and localisation using histopathology as reference: A prospective single-centre diagnostic accuracy study

Article References: Gossili, F., Harving, F., Petersen, A. C., Madsen, C., Bouchelouche, K., Bruun, N. H., Leusink, R. J., Ahmed, A., Laursen, A. M., & Zacho, H. D. (2026). Head-to-head comparison of [¹⁸F]F-PSMA-1007 PET/CT and [⁶⁸Ga]Ga-PSMA-11 PET/CT for intraprostatic tumour detection and localisation using histopathology as reference: A prospective single-centre diagnostic accuracy study. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08191-9

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08191-9

Keywords: PSMA PET, prostate cancer, PET/CT, multiparametric MRI, F-18 PSMA-1007, Ga-68 PSMA-11, histopathology, tumour localisation, diagnostic accuracy, radical prostatectomy, nuclear medicine, PI-RADS