Poster Poster Program Therapy Physics

Integrating Multimodal Radiomics with Deep Learning to Predict Survival for Patients with Brain Metastases Underwent Radiotherapy

Abstract
Purpose

Survival prediction based on radiomics and deep learning may aid treatment decision-making and follow-up management. This study aims to develop and validate an integrated model integrating multimodal imaging (MRI and CT), radiomics, and deep learning features to individually predict patients' overall survival after radiotherapy.

Methods

This multicenter retrospective study included 173 patients. Hospital A enrolled 141 patients, and Hospital B enrolled 32 patients. Features were extracted from preoperative MR and CT images using both radiomics and deep learning based on the ConvNeXt network, followed by feature selection via univariate Cox regression and LASSO regression. Unimodal and multimodal radiomics models, deep learning models, and their integrated models were constructed and evaluated on internal validation and external testing sets. Model performance was assessed using the concordance index (C-index), time-dependent area under the curve (Time-AUC), and Kaplan-Meier curves, with model interpretability analyzed via the SHAP method.

Results

The integrated model integrating multimodal radiomics and deep learning features (R_MR+DL_MR+R_CT+DL_CT) demonstrated the best predictive performance, achieving C-index values of 0.769, 0.737, and 0.709 in the training set, internal validation set, and external test set, respectively, significantly outperforming any single-modality or single-method model. The model showed superior predictive efficacy in the lung cancer subgroup and the WBRT+SIB treatment subgroup. Time-AUC analysis further confirmed the model's stable and favorable performance in predicting short-term (90 days), medium-term (180 days), and long-term (365 days) survival. SHAP analysis successfully identified key imaging features that contributed most to the prognosis.

Conclusion

The integrated model based on pretreatment CT and MR developed in this study can predict overall survival of patients with brain metastases underwent radiotherapy, which may help in risk stratification and guide treatment decision-making and follow-up management.

People

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
Python-Based Automation Framework for Annual Machine QA Data Archiving In Qatrack+

Annual water-tank measurements help ensure beam characteristics remain consistent with commissioning baselines. However, the lack of a standardized processing workflow and decentralized data storage makes it difficult to analyze...

Syed Bilal Ahmad, PhD
Therapy Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
User Expectations and Current Availability of HDR Brachytherapy Audits In Europe

The aim of this work was to evaluate the need to implement more dosimetric audits in high‐dose‐rate brachytherapy (HDR-BT) in Europe and to identify which characteristics such audits should meet according to users.

Javier Vijande, PhD Laura Oliver Cañamás
Therapy Physics 0 people interested