Radiomics artificial intelligence modelling for prediction of local control for colorectal liver metastases treated with radiotherapy

• DOI: 10.1016/j.phro.2022.09.004
• Publication Date: 2022-09-13T17:30:22Z
• Authors: Ricky Hu; Ishita Chen; Jacob Peoples; Jean-Paul Salameh; Mithat Gönen; Paul B. Romesser; Amber L. Simpson; Marsha Reyngold
• Keywords: Progression-free survival

Prognostic assessment of local therapies for colorectal liver metastases (CLM) is essential guiding management in radiation oncology. Computed tomography (CT) contains texture information which may be predictive metastatic environments. To investigate the feasibility analyzing CT texture, we sought to build an automated model predict progression-free survival using radiomics and artificial intelligence (AI).Liver scans outcomes N = 97 CLM patients treated with radiotherapy were retrospectively obtained. A was built by extracting 108 radiomic features from tumor volumes a random forest (RSF) progression. Accuracies measured concordance indices (C-index) integrated Brier scores (IBS) 4-fold cross-validation. This repeated different segmentations clinical variables as inputs RSF. Predictive identified perturbation importances.The AI achieved C-index 0.68 (CI: 0.62-0.74) IBS below 0.25 most feature gray tone difference matrix strength (importance: 1.90 CI: 0.93-2.86) treatment maximum dose 3.83, 1.05-6.62). The data only similar 0.62 0.56-0.69), suggesting that signals exist data.The good prediction accuracy CLM, providing support or combined machine learning aid prognostic management.