Hybrid Deep Learning for Survival Prediction in Brain Metastases Using Multimodal MRI and Clinical Data

Diagnostics, Год журнала: 2025, Номер 15(10), С. 1242 - 1242

Опубликована: Май 14, 2025

Background: Survival prediction in patients with brain metastases remains a major clinical challenge, where timely and individualized prognostic estimates are critical for guiding treatment strategies patient counseling. Methods: We propose novel hybrid deep learning framework that integrates volumetric MRI-derived imaging biomarkers structured demographic data to predict overall survival time. Our dataset includes 148 from three institutions, featuring expert-annotated segmentations of enhancing tumors, necrosis, peritumoral edema. Two convolutional neural network backbones-ResNet-50 EfficientNet-B0-were fused fully connected layers processing tabular data. Models were trained using mean squared error loss evaluated through stratified cross-validation an independent held-out test set. Results: The model based on EfficientNet-B0 achieved state-of-the-art performance, attaining R2 score 0.970 absolute 3.05 days the Permutation feature importance highlighted edema-to-tumor ratio tumor volume as most informative predictors. Grad-CAM visualizations confirmed model's attention anatomically clinically relevant regions. Performance consistency across validation folds framework's robustness generalizability. Conclusions: This study demonstrates multimodal can deliver accurate, explainable, actionable predictions metastases. proposed offers promising foundation integration into real-world oncology workflows support personalized prognosis informed therapeutic decision-making.

Язык: Английский

Morphometric and radiomics analysis toward the prediction of epilepsy associated with supratentorial low-grade glioma in children

Cancer Imaging, Год журнала: 2025, Номер 25(1)

Опубликована: Май 19, 2025

Язык: Английский