The research, led by Natalia Abian-Franco and F. Javier Falcó-Martí of Hospital Reina Sofía in Tudela and the Institute of Biocomputation and Physics of Complex Systems at the University of Zaragoza, together with senior authors Adrián Velazquez-Campoy and Olga Abian, combined a technique called Thermal Liquid Biopsy with machine learning to classify ovarian tumors as benign or malignant before surgery. The idea behind the approach is deceptively simple. When blood serum is heated in a controlled way using differential scanning calorimetry, the proteins and other molecules suspended in it unfold at characteristic temperatures, producing a thermal fingerprint known as a thermogram. Because cancer changes the composition of the blood, including the abundance and modification states of abundant proteins such as albumin, malignant disease leaves a detectable imprint on that fingerprint.

Differential scanning calorimetry has long been a staple of biophysics laboratories, where it is used to measure the stability of proteins and the binding of drugs to their targets. Applying it to clinical diagnostics is a more recent development, and this study is one of the most thorough attempts to push the technique toward real-world use in oncology. The team analyzed serum samples from 312 patients with a range of ovarian pathologies, drawing on a prospective cohort of 242 patients recruited between 2018 and 2021 and a retrospective cohort of 70 patients whose samples dated from 2011 to 2018. Across the pooled dataset, 186 tumors were benign and 126 were malignant, a distribution that reflects the genuine clinical challenge: in a typical gynecology service, benign masses outnumber malignant ones, which makes it easy for a diagnostic test to look good simply by defaulting to the common answer.

That class imbalance shaped the design of the study in an important way. The researchers built two predictive models. The first, called the OP-only model, was trained exclusively on the prospective samples, giving the most honest and conservative estimate of how the approach would perform on newly recruited patients. The second, an exploratory enriched model designated OP plus OV, added the retrospective samples to the training data, deliberately enriching the malignant class to see whether more cancer examples would sharpen the algorithm’s discrimination. Both models integrated features extracted from the thermograms with clinical and biochemical variables, including established markers such as CA125 and HE4, into what the authors call an integrated TLB and clinical model, or iTLB plus iClin.

The results were striking. The conservative model, trained only on prospective data, achieved an area under the receiver operating characteristic curve of 0.85 for detecting malignancy. In practical terms, an AUC of 0.85 means the model distinguished benign from malignant tumors substantially better than chance and better than many existing single-marker approaches, though it still leaves room for error in individual cases. The enriched model, combining prospective and retrospective data, reached a numerically higher AUC of 0.91 and showed improved classification of indeterminate cases, the gray zone where imaging and biomarkers alone leave clinicians genuinely uncertain. The authors are careful, however, to flag a critical caveat: the retrospective cohort contained malignant cases only, which means the enriched model’s apparent advantage could partly reflect the composition of its training data rather than a genuine leap in diagnostic power.

This kind of methodological honesty matters, because diagnostic studies in oncology have a long history of promising results that fade on contact with clinical reality. The Zaragoza team took several steps to guard against the most common failure modes. They evaluated performance using nested cross-validation, a rigorous scheme in which model tuning and performance estimation are kept strictly separate, reducing the risk of optimistic bias. They also applied regularization techniques, which constrain the complexity of the machine learning models and reduce overfitting, the phenomenon in which an algorithm memorizes the quirks of its training set rather than learning generalizable patterns. Even so, the authors state plainly that external validation in independent, multicenter cohorts is a mandatory prerequisite before any clinical implementation can be considered.

The appeal of Thermal Liquid Biopsy lies partly in what it does not require. Conventional liquid biopsies hunt for tumor DNA, circulating tumor cells, or specific molecular markers, all of which demand that the tumor shed identifiable material into the bloodstream in detectable quantities. Thermal profiling, by contrast, reads a global physical property of the serum, capturing the aggregate consequence of the many molecular changes that cancer induces rather than betting on any single molecule. That makes it conceptually robust to the heterogeneity that makes ovarian cancer so difficult: different tumors, and even different regions of the same tumor, can drive malignancy through different molecular routes, yet all of them perturb the serum proteome in ways that shift the thermal curve. Combined with clinical variables such as age, body mass index, and standard biomarkers, the thermogram gives the machine learning algorithm a rich, multidimensional picture of each patient.

The clinical stakes of better preoperative classification are considerable. When a woman presents with an ovarian mass, the surgical plan diverges sharply depending on the suspected diagnosis. A likely benign mass may be handled by a general gynecologist with a straightforward, fertility-preserving procedure, while a likely malignancy calls for a gynecologic oncologist, often with full surgical staging, and ideally at a specialized cancer center. Misclassification in either direction is costly: benign tumors overtreated with radical surgery expose patients to unnecessary morbidity, while cancers understaged at a first operation can require repeat surgery and may compromise oncological outcomes. Current tools, including ultrasound scoring systems and biomarkers such as CA125 and the risk of malignancy index, leave a meaningful fraction of cases indeterminate, which is precisely where the enriched model in this study showed its most interesting gains.

There are also practical advantages in cost and workflow. Differential scanning calorimetry instruments are standard equipment in biophysics labs, the assay requires only a small volume of serum, and the measurement itself is fast and label-free, requiring no antibodies, amplification chemistry, or sequencing. Turning raw thermograms into diagnostic predictions does require computational infrastructure and validated models, but that is a solvable engineering problem once the clinical validity is established. The Spanish group’s work, funded by the Spanish Ministry of Science, Innovation and Universities, the Instituto de Salud Carlos III, and regional and charitable funders, was conducted with institutional ethics approval and written informed consent from all participants, with samples supplied through the Navarrabiomed Biobank under a formal transfer agreement.

For now, the message from the study is one of cautious optimism. The conservative AUC of 0.85 from purely prospective data establishes that thermal fingerprints carry real diagnostic information about ovarian tumors, and the exploratory results hint that larger, better-balanced training sets could push performance higher. But the path from a promising single-center model to a tool used in operating theaters runs through the unglamorous work of multicenter validation, standardization of sample handling, and prospective trials in genuinely independent populations. If those steps succeed, the idea that a simple heating curve drawn from a tube of blood, interpreted by an algorithm, could help decide how a woman’s cancer surgery is planned may move from biophysics curiosity to standard of care, and ovarian cancer’s devastating pattern of late detection may finally acquire a new line of defense.

Subject of Research: Machine learning analysis of serum thermal profiles for preoperative diagnosis of malignant ovarian tumors

Article Title: Integration of machine learning with thermal liquid biopsy for preoperative assessment of malignant ovarian tumors

Article References: Abian-Franco, N., Falcó-Martí, F. J., Hermoso-Durán, S., Sánchez-Gracia, O., Vega, S., Ortega-Alarcón, D., Velazquez-Campoy, A., & Abian, O. (2026). Integration of machine learning with thermal liquid biopsy for preoperative assessment of malignant ovarian tumors. BMC Medicine. https://doi.org/10.1186/s12916-026-05212-0

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05212-0

Keywords: ovarian cancer, thermal liquid biopsy, machine learning, differential scanning calorimetry, liquid biopsy, serum biomarkers, preoperative diagnosis, CA125, HE4, risk stratification, BMC Medicine, predictive medicine

Researchers led by Enlong Zhang and Yuan Li of Peking University Third Hospital, together with colleagues at Tsinghua University Hospital and Peking University International Hospital, developed a tumor habitat-based radiomics approach to distinguish retroperitoneal non-fatty dedifferentiated liposarcoma from leiomyosarcoma. Their retrospective study included 166 patients, 102 with dedifferentiated liposarcoma and 64 with leiomyosarcoma, all of whom underwent multiphase contrast-enhanced CT. Rather than treating each tumor as a single uniform blob, the team borrowed a concept from oncology known as intratumoral heterogeneity, the recognition that a cancer is a patchwork of biologically distinct microenvironments, each with its own cellularity, vascularity, necrosis, and stromal composition.

The technical core of the method is elegantly simple. On the largest axial slice of each tumor, the researchers applied a Simple Linear Iterative Clustering algorithm, a superpixel segmentation technique originally developed for computer vision, to divide the tumor image into hundreds of small, coherent regions of similar pixel intensity. A K-means clustering step then grouped these superpixels into three habitat subregions, effectively partitioning each tumor into three zones that reflect different levels of attenuation, a radiological proxy for tissue density. The team designated these subregions ROI1, ROI2, and ROI3, with ROI2 corresponding to the intermediate-attenuation zone that often represents a transitional territory between viable cellular tissue and degenerative change.

From each habitat subregion, across three imaging phases, the non-contrast phase, the arterial phase, and the venous phase, the researchers extracted a comprehensive panel of quantitative radiomics features. These included shape descriptors such as major axis length, first-order statistics capturing the distribution of gray values, and higher-order texture features derived from gray-level co-occurrence matrices, run-length matrices, size zone matrices, dependence matrices, and neighborhood gray-tone difference matrices, as well as wavelet-transformed versions of these features that probe texture at multiple spatial scales. Feature selection was performed independently within each subregion using LASSO regression, and predictive models were then built with XGBoost, a gradient-boosted decision tree algorithm renowned for its performance on tabular biomedical data.

The results point to a clear conclusion: contrast matters. Habitat radiomics models derived from contrast-enhanced CT generally outperformed those built on non-contrast images, suggesting that the way tumor subregions take up iodinated contrast, and therefore their vascular perfusion characteristics, carries diagnostic information that plain density measurements miss. The single best-performing model was built from the venous-phase intermediate-attenuation subregion, VP-ROI2. In the testing cohort it achieved an area under the receiver operating characteristic curve of 0.834, with a 95 percent confidence interval of 0.713 to 0.950, alongside an accuracy of 0.820, a sensitivity of 0.871, and a specificity of 0.737.

The venous-phase advantage is physiologically plausible. The venous phase of a contrast-enhanced CT, typically acquired around sixty to ninety seconds after injection, captures the period of maximum enhancement in many soft-tissue tumors, when contrast has diffused into the extracellular space and reflects capillary permeability and interstitial volume. Dedifferentiated liposarcoma and leiomyosarcoma differ in their microvascular architecture and stromal composition, and those differences apparently imprint distinguishable enhancement patterns on the intermediate-attenuation habitat. Interestingly, VP-ROI2 showed only a nominally significant advantage over its non-contrast counterpart, NP-ROI2, with a P value of 0.048 that fell to an adjusted P value of 0.173 after false discovery rate correction, a statistical caution the authors appropriately acknowledge. ROI2 nevertheless emerged as the most discriminative subregion on both arterial and venous phases.

Perhaps the most compelling aspect of the study is its commitment to interpretability, a virtue not always found in machine learning medicine. Using SHAP, or SHapley Additive exPlanations, a game-theoretic framework that quantifies each feature’s contribution to individual predictions, the team identified original_shape_MajorAxisLength and wavelet-based texture features as the most influential predictors. Shape matters, which makes intuitive sense: dedifferentiated liposarcomas and leiomyosarcomas differ in their growth patterns and margins. But the prominence of wavelet texture features, which encode fine-scale spatial patterning of pixel intensities, indicates that the microscopic organization of tumor tissue, its heterogeneity at radiologically invisible scales, is what truly separates the two diagnoses. The models were built and reported following established standards, including the Image Biomarker Standardization Initiative nomenclature and the TRIPOD guidelines for prediction model reporting, and the study received an assessment of radiomics quality.

The clinical implications are significant. Today, distinguishing these sarcomas preoperatively often depends on biopsy, and even then core needle sampling of a large heterogeneous tumor can miss the diagnostic territory, a problem known as sampling error. Accurate preoperative differentiation guides the surgical plan: dedifferentiated liposarcoma demands wide clearance of the fatty component along with the dedifferentiated nodule, whereas leiomyosarcoma management hinges more on vascular involvement and multimodal planning. A noninvasive imaging biomarker that can flag the likely diagnosis before the first incision could help surgeons counsel patients, select neoadjuvant strategies, and design clinical trials that stratify patients by histological subtype rather than by the blunt label of retroperitoneal sarcoma.

The authors are careful to frame this as a promising step rather than a finished clinical tool. The cohort was retrospective and drawn from Chinese hospitals, the tumors were segmented on the largest axial slice rather than in three dimensions, and the performance figures, while respectable, leave room for improvement before the model could stand beside pathology. External validation in independent, multiethnic cohorts and prospective testing will be essential before VP-ROI2-based habitat radiomics enters routine decision-making. Still, the study offers a vivid demonstration of a broader idea now sweeping through oncological imaging: that every tumor is a landscape of habitats, and that the map of that landscape, drawn automatically from a routine contrast-enhanced CT scan, may hold answers that biopsies, radiologists, and even pathologists struggle to provide. The venous phase, long treated as a routine addendum to the arterial spectacle, turns out to be where the sarcomas reveal their secrets.

Subject of Research: Venous-phase CT-based tumor habitat radiomics for differentiating retroperitoneal non-fatty dedifferentiated liposarcoma from leiomyosarcoma

Article Title: Venous-phase CT-based tumor habitat radiomics for differentiating retroperitoneal non-fatty dedifferentiated liposarcoma from leiomyosarcoma

Article References: Zhang, E., Li, Y., Ma, L., Ji, D., Zhang, M., & Lang, N. (2026). Venous-phase CT-based tumor habitat radiomics for differentiating retroperitoneal non-fatty dedifferentiated liposarcoma from leiomyosarcoma. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02784-4

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02784-4

Keywords: retroperitoneal sarcoma, dedifferentiated liposarcoma, leiomyosarcoma, radiomics, tumor habitat, intratumoral heterogeneity, computed tomography, machine learning, XGBoost, SHAP, cancer imaging, preoperative diagnosis