AI in Oncology
Research and news where artificial intelligence is the method — diagnosis, drug discovery, pathology, clinical decision support.
Development and validation of an integrated intratumoral-peritumoral habitat radiomics model on multiparametric MRI for predicting biochemical recurrence in patients after radical prostatectomy
Ji YT, et al
A new MRI-based combined clinical habitat model (CHPM) integrating intra- and peritumoral radiomics features significantly outperformed the established CAPRA-S scoring system for predicting short-term biochemical recurrence after radical prostatectomy for prostate cancer. Developed and validated in …
Source →Multi-Sequence Fusion MRI Deep Learning Models for Discrimination of Recurrent Nasopharyngeal Carcinoma and Osteoradionecrosis.
Gong W, et al
A dual-branch 3D-ResNet-50 deep learning model combining two MRI sequences achieved an AUC of 0.86 and accuracy of 84% — with sensitivity of 91%, specificity of 72%, and an F1 score of 0.87 — in distinguishing recurrent nasopharyngeal carcinoma (rNPC) from osteoradionecrosis (ORN) after radiotherapy…
Source →Deep Learning Pipeline for Automatic Segmentation, Classification, and Molecular Subtyping of Three Pediatric Posterior Fossa Tumors Using T2-Weighted MRI.
Jin Y, et al
A fully automated deep learning pipeline using standard T2-weighted MRI has been developed and validated to segment, classify, and molecularly subtype three pediatric posterior fossa tumors — medulloblastoma, ependymoma, and pilocytic astrocytoma — achieving tumor segmentation Dice scores of 0.94–0.…
Source →Artificial Intelligence in equivocal prostatic needle biopsies: benign mimics and cancer detection.
Bawahab AA
A structured narrative review synthesizing evidence from PubMed/MEDLINE, Scopus, Web of Science, and Google Scholar through May 15, 2026 concludes that AI applied to digitized hematoxylin-and-eosin slides can support cancer detection, grading, workflow prioritization, and quality assurance in diagno…
Source →Segmentation-guided peritumoral radiomics and machine learning for ultrasound-based classification of salivary gland tumors.
Chen H, et al
A segmentation-guided machine learning pipeline combining YOLOv8 automated tumor segmentation with a Support Vector Classifier achieved area-under-the-curve scores of 0.8117 and 0.7951 in an independent testing cohort (n = 121) and an internal validation cohort (n = 51), respectively, for distinguis…
Source →Diagnostic performance of radiomic features from FS-T2WI for parotid tumor classification: A comparison of deep learning and machine learning models.
Mao X, et al
A deep learning model called TabResNet achieved the highest diagnostic accuracy for distinguishing benign from malignant parotid tumors on MRI, reaching an area under the ROC curve (AUC) of 0.8913 — outperforming the best conventional machine learning model, Random Forest (AUC = 0.8641). The study a…
Source →Thymic Radiation is Associated with Worse Outcomes in Patients with NSCLC.
Prudente V, et al
A multicohort study of 1,107 non-small cell lung cancer (NSCLC) patients has found that incidental radiation exposure to the thymus during radiotherapy is independently associated with significantly higher risks of distant metastasis and death, with each 1 Gy increase in Mean Thymic Dose linked to a…
Source →Explainable Deep Learning of Transcriptomes Prioritizes Candidate Biomarkers With Preferential Performance in Gastric Cardia Cancer Cohorts.
Xu J, et al
An explainable deep learning classifier trained on public transcriptome data achieved a near-perfect cross-validated AUC of 1.00 (both ROC and precision-recall) for distinguishing gastric cardia tumors from normal tissue, and demonstrated preferentially stronger external performance in cardia-specif…
Source →Machine Learning-Derived Immune Gene Signature Predicts Prognosis and Therapeutic Vulnerabilities in Multiple Myeloma.
Wang K, et al
Researchers developed a 12-gene immune-related gene signature (IRGS) using an integrative machine-learning framework that consistently stratified overall survival in multiple myeloma patients across multiple independent cohorts, with prognostic discrimination comparable to IFM15 and generally exceed…
Source →Ultrasound Habitat Radiomics for Preoperative Prediction of Invasive Breast Cancer With a DCIS Component: A Dual-center Retrospective Study.
Fan L, et al
An ultrasound-based habitat subregional radiomics model achieved an area under the ROC curve (AUC) of 0.910 (95% CI: 0.883–0.936) when predicting preoperatively whether a breast cancer harbors a concurrent ductal carcinoma in situ (DCIS) component, outperforming both standard imaging and single-regi…
Source →This page is educational — it is not medical advice and does not replace consultation with an oncologist. Diagnostic and treatment decisions are made solely by specialist physicians.