To develop and validate an interpretable machine learning (ML) framework that differentiates adrenal lipid-poor adenomas (LPAs) from pheochromocytomas (PHEOs) using radiomic features derived from multi-phase contrast-enhanced computed tomography (CECT).
This retrospective study included 229 patients with pathologically confirmed adrenal tumours (132 LPAs, 97 PHEOs). Lesions were stratified into washout (absolute percentage washout [APW] > 60%) and non-washout (APW ≤ 60%) cohorts according to established criteria. We trained decision tree (DT), support vector machine (SVM), and logistic regression (LR) models using a predefined set of clinically relevant CT features. The inherently interpretable DT model was further dissected to revealr key discriminative features and its underlying decision logic. Model performance was evaluated using the area under the curve (AUC) with 95% confidence intervals.
The DT model achieved excellent diagnostic performance, with an AUC of 0.977 (95% CI: 0.934–1.000) in the washout group and 0.983 (95% CI: 0.962–1.000) in the non-washout group. Feature importance analysis identified cystic degeneration as the most influential discriminator, followed by enhancement potential (EP) and baseline attenuation (CTu). The DT’s hierarchical decision rules provided clear and clinically transparent pathways.
This interpretable ML framework accurately distinguishes LPAs from PHEOs. The DT model strikes an optimal balance between high diagnostic accuracy and inherent interpretability, offering a practical tool that may enhance clinical confidence when managing indeterminate adrenal lesions.
Pan-Liang Zhao, F. Zhu, C. Yan et al.· Frontiers in Artificial Inte...· 0 citations
PURPOSE
Cancer-associated fibroblasts (CAFs) are central drivers of PDAC progression and therapeutic resistance, yet their preoperative clinical utility remains unexplored. We aimed to translate single-cell CAF heterogeneity into a CT-based framework for preoperative risk stratification and adjuvant chemotherapy stratification in PDAC.
MATERIALS AND METHODS
A total of 1452 PDAC patients were included across transcriptomic and imaging analyses. Single-cell RNA sequencing data from 24 patients were integrated with TCGA-PAAD bulk transcriptomics using the Scissor algorithm to identify prognosis-associated CAF subpopulations. A nine-gene CAPR score was validated across four independent transcriptomic cohorts (n = 594). A CT-based topological data analysis (TDA) classifier (ra-CAPR) was developed in an institutional cohort (n = 122), externally validated in the TCIA cohort (n = 50), and evaluated in an independent surgical cohort (n = 657), with performance benchmarked against conventional radiomics.
RESULTS
Three adverse CAF subpopulations (ECM-remodelling myCAF, hypoxic CAF, and iCAF_chemokine) were identified, defining an immune-excluded, mutationally-burdened, and chemoresistant phenotype across validation cohorts. The TDA-based ra-CAPR classifier demonstrated superior cross-cohort robustness over conventional radiomics (AUC 0.774 vs. 0.715; 0.744 vs. 0.621), accompanied by markedly lower cross-cohort feature distributional shift (26% vs. 75%). In the surgical cohort, ra-CAPR independently predicted overall survival (HR 1.42, P = 0.005). Adjuvant chemotherapy improved OS in both ra-CAPR subgroups, with a larger absolute gain in ra-CAPR-low patients (29.5 vs 17.0 months).
CONCLUSION
By translating adverse CAF subpopulation signatures into a CT-based topological biomarker, ra-CAPR enables individualized preoperative risk stratification and may facilitate risk-adapted postoperative management.
Lingling Wang, Zongyu Xie, Qiying Tang et al.· European Journal of Radiolog...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
