Across the complete renal cell carcinoma cohort, the CAIX-targeted technique detected 90.9% of the 492 true lesions, while FDG detected 56.7%. The difference was more pronounced in the clear cell renal cell carcinoma subgroup, which accounted for the study’s strongest results: lesion detection reached 97.9% with [⁶⁸Ga]Ga-DPI-4452 and 47.1% with FDG. Patient-level diagnostic accuracy in clear cell disease was 98.2% for the CAIX-targeted technique and 77.2% for ¹⁸F-FDG.

The imaging strategy targets carbonic anhydrase IX, commonly abbreviated CAIX, a protein that the researchers described as being highly expressed in clear cell renal cell carcinoma but minimally expressed in normal renal tissue. Lead author Professor Esra Arslan said: “CAIX gives us a highly selective biological target. It is expressed at a very high rate in ccRCC, while expression in normal renal tissue remains minimal.

This gives us the potential to distinguish tumour from surrounding healthy tissue much more clearly.” The target therefore offers a way to identify clear cell tumours without relying on the glucose metabolism measured by ¹⁸F-FDG. FDG PET/CT measures glucose metabolism and is commonly used in cancer imaging, but its usefulness in renal cell carcinoma can be limited by high physiological glucose uptake and intense urinary excretion in the kidneys. These processes can make tumours harder to distinguish from surrounding tissue.

Some low-grade tumours also take up relatively little FDG, further complicating their detection. In the prospective comparison, [⁶⁸Ga]Ga-DPI-4452 was designed to target CAIX rather than glucose activity, allowing the investigators to test whether greater selectivity for clear cell renal cell carcinoma would produce more conspicuous lesions and improved detection in this cohort. Quantitative measurements confirmed marked differences in tumour visibility.

Uptake of [⁶⁸Ga]Ga-DPI-4452 in tumours was approximately five times higher than uptake of ¹⁸F-FDG, while the tumour-to-background ratio was around ten times greater overall. The tumour-to-background ratio describes how clearly a tumour stands out from surrounding tissue. In identified brain metastases, for example, the median maximum tumour-to-background ratio was 281.3 for [⁶⁸Ga]Ga-DPI-4452, compared with 0.7 for ¹⁸F-FDG.

These measurements showed that the CAIX-targeted agent generated substantially stronger contrast in the lesions assessed, although they do not by themselves establish better long-term clinical outcomes. Detection differences were also evident at specific metastatic sites. The CAIX-targeted technique detected all identified brain metastases and all identified lung metastases, while FDG detected around half of the lung metastases.

For lymph nodes, detection was 95.0% with [⁶⁸Ga]Ga-DPI-4452 and 68.9% with ¹⁸F-FDG. Brain metastases provided a particularly clear example of the contrast difference because high physiological FDG uptake can make lesions difficult to distinguish. These site-specific findings apply to the lesions identified and evaluated in the study and do not establish equivalent performance for every form or stage of metastatic renal cell carcinoma.

Liver metastases were the main exception in the overall cohort. FDG performed better for liver lesions, which the researchers linked to lower and more variable CAIX expression in some non-clear cell renal cell carcinoma subtypes. Within the clear cell subgroup, however, performance again favoured [⁶⁸Ga]Ga-DPI-4452.

The variation is important for interpreting the headline result: the near-complete detection reported for clear cell renal cell carcinoma should not be treated as a universal performance estimate for renal cell carcinoma. The broader cohort results and the liver finding also show why future studies must account for biological differences among subtypes. The investigators examined whether the imaging results affected decisions about disease stage and treatment.

A change in clinical management or disease stage occurred in 30.9% of patients assessed with [⁶⁸Ga]Ga-DPI-4452, compared with 5.9% after ¹⁸F-FDG. Reported changes included identifying previously undetected recurrent or residual disease and detecting nodal or distant metastases that resulted in upstaging or reclassified some patients from oligometastatic, or limited, disease to more widespread metastatic disease. In three patients, findings considered positive on FDG PET/CT were excluded by the CAIX-targeted examination, which the researchers reported had avoided unnecessary interventions.

The study measured imaging and management changes, not whether those changes ultimately improved survival or other patient outcomes. Arslan and Professor Dr Halil Kömek, who contributed equally to the work, identified possible roles for CAIX-targeted PET/CT in initial staging before nephrectomy or metastasectomy, restaging when recurrence is suspected and, in the future, selection for treatment with a matched radioligand. The proposed theranostic strategy would use [⁶⁸Ga]Ga-DPI-4452 imaging to identify patients who might be suitable for treatment with [¹⁷⁷Lu]Lu-DPI-4452.

That application remains investigational in the congress report. Arslan said [⁶⁸Ga]Ga-DPI-4452 “has the potential to become a valuable tool across several stages of RCC management,” but cautioned that it was premature to call the agent a new molecular imaging standard without validation in larger multicentre studies stratified by subtype. The prospective design and direct comparison strengthen the reported evidence, but several limitations remain.

Longer prospective studies will be needed to determine whether the observed changes in staging and clinical management translate into improved patient outcomes.

Subject of Research: Prospective evaluation of CAIX-targeted PET/CT for detecting renal cell carcinoma lesions in 68 patients.

Article Title: New PET imaging approach detects almost all lesions in common form of kidney cancer

Article References: New PET imaging approach detects almost all lesions in common form of kidney cancer. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: CAIX-targeted PET/CT, clear cell renal cell carcinoma, kidney cancer, positron emission tomography, ¹⁸F-FDG PET/CT, carbonic anhydrase IX, lesion detection, cancer staging, metastases, EANM 2026, imaging, approach

The research, led by Hai-Tao Zhu, Yan-Jie Shi, Xiao-Ting Li and Ying-Shi Sun, enrolled 37 patients with breast cancer who harbored a total of 191 spinal metastatic lesions. Every patient underwent diffusion weighted imaging, or DWI, before chemotherapy started. DWI is a magnetic resonance technique that measures how freely water molecules move within tissue. In tumors, dense cell packing restricts water motion, and this restriction can be quantified. The team then split the patients into an investigating group of 19 and an independent test group of 18, a design that allowed them to build the model on one cohort and validate it on data the model had never seen.

The technical innovation at the heart of the study lies in how the imaging data were processed. Traditional DWI analysis acquires images at multiple b-values, which are gradient-strength settings that determine how sensitive each image is to water diffusion. Researchers typically fit the signal decay across these b-values into a predefined mathematical model, such as the apparent diffusion coefficient, the intravoxel incoherent motion model, the diffusion kurtosis model, or the stretched exponential model. Each model yields parameters like ADC, DDC, α, D*, D_true, frac, D_app and K_app, each intended to capture a different aspect of tissue microstructure. The problem is that these models impose assumptions on the data, and if the underlying biology does not match the assumed curve shape, information is lost or distorted.

Zhu and colleagues took a different route. Instead of fitting the multi-b-value signals into any model, they converted the raw signal intensities within each lesion into histograms, one histogram per b-value, and then concatenated all ten histograms into a single composite descriptor they call a histogram signature. This signature preserves the full distribution of signal intensities across the entire b-value range without forcing the data into a curve-fitting framework. Two radiologists manually delineated the borders of each lesion on the pretreatment DWI images to define the regions of interest, ensuring that the signatures were derived from tumor tissue rather than surrounding bone marrow or muscle.

The histogram signatures were then fed into a convolutional neural network, a class of deep-learning architecture that excels at detecting spatial and statistical patterns in structured data. The network’s task was binary classification: given a pretreatment signature, predict whether the corresponding patient’s disease would progress under chemotherapy, designated progressive disease or PD, or fall into the non-PD category. The network was trained on the investigating group and evaluated on the test group using receiver operating characteristic analysis, the standard method for quantifying diagnostic discrimination.

The results were striking. The histogram-signature model achieved an area under the ROC curve, or AUC, of 0.882, with a 95 percent confidence interval spanning 0.803 to 0.938. An AUC of 0.882 indicates strong discriminatory power, meaning the model could correctly rank a randomly chosen progressive-disease patient above a randomly chosen responder roughly 88 percent of the time. By comparison, the best-performing individual fitted parameter, D_app, reached an AUC of only 0.685, with a confidence interval of 0.585 to 0.774. The difference between the two approaches was statistically significant, with a Z statistic of 3.312 and a P value of 0.0009, confirming that the advantage of the histogram signature over the conventional parameter was not a chance finding.

Why does the model-free approach outperform the fitted parameters? The answer likely lies in what each method captures. A single fitted parameter like ADC or D_app compresses the entire diffusion signal decay into one number per voxel, discarding information about the shape of the decay curve, the heterogeneity within the tumor, and the distribution of signal values across the lesion. Histogram signatures, by contrast, retain the full statistical profile at every b-value. Tumors are notoriously heterogeneous, containing regions of necrosis, densely packed viable cells, and variable vascularity, all of which influence diffusion behavior in ways that a single scalar parameter cannot fully represent. By preserving this complexity, the signature gives the neural network a richer substrate from which to learn the patterns that distinguish responders from non-responders.

The clinical implications are considerable. Spinal metastases from breast cancer are common and often painful, and progressive disease in the spine can lead to vertebral collapse, spinal cord compression, and severe neurological impairment. If oncologists could identify, before the first cycle of chemotherapy, which patients are unlikely to respond, they could escalate to alternative regimens, add local therapies such as radiotherapy or ablation, or intensify surveillance for skeletal complications. Conversely, patients predicted to respond could proceed with confidence, avoiding unnecessary changes to a working plan. The pretreatment nature of the prediction is key: it moves decision-making from a reactive posture, waiting weeks for follow-up imaging to reveal whether the tumor is shrinking, to a proactive one grounded in data available on day one.

The study is not without limitations. Thirty-seven patients and 191 lesions constitute a modest sample size, and the retrospective design means the model was built and tested on historical data rather than prospectively enrolled patients. The radiologists’ manual delineation of lesion borders, while carefully performed, introduces a degree of operator dependence that automated segmentation could eventually reduce. External validation at other institutions, with different MRI scanners and imaging protocols, will be essential before the approach can be considered ready for routine clinical deployment. The authors note that the work was supported by the National Key Research and Development Program of China under grant 2023YFC3402805, and the study received ethics approval from Peking University Cancer Hospital with informed consent obtained from all enrolled subjects.

Nevertheless, the findings point toward a broader shift in oncological imaging. Rather than asking radiologists to extract a handful of predefined parameters and interpret them through clinical experience, the field is moving toward models that learn directly from the full richness of the imaging data. The histogram-signature approach demonstrated here sidesteps the long-standing tension between mathematical elegance and biological reality, showing that sometimes the best way to model diffusion is not to model it at all. If future prospective studies confirm these results, the combination of multi-b-value DWI and deep learning could become a standard pretreatment tool for patients facing spinal metastases, offering a data-driven answer to the question that matters most: will this chemotherapy work for me?

Subject of Research: Pretreatment prediction of the response to chemotherapy for spine metastases of breast cancer using histogram signatures from diffusion weighted imaging with multiple b-values: a retrospective study

Article Title: Pretreatment prediction of the response to chemotherapy for spine metastases of breast cancer using histogram signatures from diffusion weighted imaging with multiple b-values: a retrospective study

Article References: Zhu, H.-T., Shi, Y.-J., Li, X.-T., & Sun, Y.-S. (2026). Pretreatment prediction of the response to chemotherapy for spine metastases of breast cancer using histogram signatures from diffusion weighted imaging with multiple b-values: a retrospective study. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02767-5

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02767-5

Keywords: Pretreatment, prediction, response, chemotherapy, spine, metastases, breast, cancer, histogram, signatures, diffusion, weighted

The research, published in CVIR Oncology by a team led by Maria Tækker of Aarhus University Hospital, examined percutaneous cryoablation, a technique in which thin, needle-like probes are inserted through the skin directly into a tumor under computed tomography guidance. Once the probes are in position, pressurized argon gas is circulated through their tips, dropping the temperature at the tumor to at least minus 40 degrees Celsius. At that extreme cold, ice crystals form inside and around cancer cells, rupturing membranes and cutting off the blood supply, ultimately killing the tissue through a process of controlled necrosis. Because the expanding ice ball is visible on CT scans as it grows, the radiologist can watch the frozen zone engulf the tumor in real time, ensuring the lethal cold extends at least five millimeters beyond the tumor’s edge while sparing nearby healthy structures.

This real-time visibility is one of cryoablation’s most significant advantages over competing ablative technologies, and it carries a second benefit that is easy to overlook: the procedure can be repeated. Unlike stereotactic body radiation therapy, where cumulative radiation dose limits how many times a region can be treated, or surgery, where each operation adds scar tissue and risk, cryoablation leaves tissue in a state that can safely tolerate another freeze cycle if a tumor returns. For patients with oligometastatic disease, the intermediate state between localized and widespread cancer in which only a limited number of metastases exist in a few organ sites, that repeatability could prove decisive.

The Danish team set out to test exactly that proposition in a group of patients who had run out of conventional options. Between December 2016 and February 2022, sixteen patients with fewer than three histologically verified metastases each were selected for treatment at a single center. Crucially, every case was reviewed by a multidisciplinary team of oncologists, surgeons, pathologists, and radiologists, and all of the patients had already completed treatment for their primary cancers. Many carried significant comorbidities: the most common were hypertension and diabetes, and the average score on the American Society of Anesthesiologists physical status scale, which grades surgical risk, was 2.6, indicating severe systemic disease in several participants. These were not healthy patients; they were people whose age or health made surgery, additional radiation, or more chemotherapy too dangerous, leaving cryoablation as the only feasible curative-intent option.

The technical execution of the procedures reflects the maturity the field has reached. A senior radiologist with a decade of experience in CT-guided interventions performed the ablations, most of them, 73.7 percent, under conscious sedation rather than general anesthesia. Using an argon-based cryoablation system with sealed 14- or 17-gauge cryoprobes, the team created ice balls that covered each tumor with a margin of at least five millimeters. Standard treatments used a double freeze-thaw cycle of ten minutes of freezing followed by eight minutes of thawing, a rhythm that exploits the way repeated freezing damages cells more thoroughly than a single cycle. Lymph node metastases received a triple freeze-thaw protocol, and when tumors sat close to vulnerable structures, the team used hydrodisplacement, injecting fluid to physically push organs such as the bowel away from the freezing zone.

The results, drawn from fifteen patients and sixteen metastases with complete follow-up, are striking for a cohort this fragile. The median tumor size was 18 millimeters, though lesions ranged from 6 to 60 millimeters, and the primary cancers originated in the kidney, lung, and breast. After a single cryoablation treatment, 80 percent of patients achieved local tumor control at a median follow-up of 23.6 months. Three patients whose tumors progressed underwent repeat ablations, and after these retreatments the so-called secondary efficacy rate rose to 93 percent at a median follow-up of 28.7 months. In other words, even when the first freeze failed to eradicate a lesion permanently, freezing it again usually finished the job, validating the technique’s key selling point of repeatability.

The safety profile was equally encouraging. Ninety percent of patients went home the same day. Of the nineteen treatments performed, only two produced post-procedural complications, and just one patient needed overnight hospitalization for monitoring. Graded on the Common Terminology Criteria for Adverse Events scale, complications were limited to grade 1 in 16 percent of procedures and grade 2 in 5 percent, with nothing more severe, and no complication required further intervention. For a population of comorbid patients with stage IV cancer, avoiding the morbidity of surgery or the accumulating toxicity of repeat radiation while still achieving local tumor control represents a genuine clinical gain.

Survival figures rounded out the picture. At one year, 93 percent of patients were alive; at two years, 72 percent; and at three years, 63 percent. The deaths that occurred were attributed to progression of the primary cancers rather than to the procedure itself. These numbers sit comfortably within the range reported by comparable studies, which have shown two-year overall survival of 82 to 94 percent and three-year survival of roughly 62 to 82 percent for metastasis-directed cryoablation. A recent retrospective analysis of 511 patients with lung metastases by Gits and colleagues found no significant survival differences between sublobar resection, stereotactic body radiation therapy, and percutaneous ablation, suggesting that needle-based approaches can stand alongside established local therapies. Meanwhile, literature on radiation-directed treatment of oligometastases indicates that such metastasis-directed therapy may extend expected survival by up to fifteen months, underscoring the stakes of getting local control right.

The study’s authors are candid about its limits. With only fifteen analyzed patients, a retrospective design, and a single center, the findings cannot be generalized blindly, and the inclusion of patients with advanced cancer and high anesthetic risk scores may actually understate the survival that healthier candidates could expect. Tumor size also emerged as a question mark: prior research suggests lesions larger than 20 millimeters carry a higher risk of local progression after cryoablation, and while this study’s tumors ranged well beyond that threshold, the sample was too small to settle the matter. What the data do support is a consistent pattern across diverse anatomical sites, from kidneys and adrenal glands to retroperitoneal soft tissue, lymph nodes, and even the iliac bone, that percutaneous cryoablation is an effective and repeatable method for managing limited metastatic disease over at least a two-year horizon.

The broader significance of the work lies in what it offers patients who have been told their options are exhausted. For someone with two or three metastases and a heart too weak for surgery, a same-day procedure that freezes their tumors under light sedation, with a 93 percent chance of lasting local control after retreatment and no serious complications, changes the calculus of stage IV cancer. The researchers argue that future clinical trials should evaluate overall survival together with quality of life, since both metrics capture what matters most when life is extended at advanced disease stages. High-level randomized evidence comparing cryoablation directly with radiation therapy for oligometastatic disease is still missing, but studies like this one, built on careful multidisciplinary selection, standardized reporting, and honest accounting of failures, are laying the groundwork. For now, the message is simple and quietly revolutionary: sometimes the best answer to a spreading cancer is not more drugs or bigger operations, but a precisely guided blast of cold, delivered through a needle, and repeated as often as needed.

Subject of Research: Percutaneous cryoablation as metastasis-directed therapy for oligometastatic cancer

Article Title: Percutaneous imaging-guided cryoablation of oligometastases – a single-center experience

Article References: Tækker, M., Pietersen, P. I., Junker, T., Jensen, C. G., Vogsen, M., Rasmussen, B. S., Welch, B. T., Persson, G. F., Al-Mashhadi, R., & Graumann, O. (2025). Percutaneous imaging-guided cryoablation of oligometastases – a single-center experience. CVIR Oncology, 1(1), Article 11. https://doi.org/10.1007/s44343-025-00011-5

Image Credits: AI Generated

DOI: 10.1007/s44343-025-00011-5

Keywords: cryoablation, oligometastatic disease, metastases, interventional radiology, CT guidance, cancer treatment, local tumor control, overall survival, minimally invasive therapy, kidney cancer, lung cancer, breast cancer