REAL-TIME MACHINE LEARNING PREDICTION OF ACUTE KIDNEY INJURY IN ICU CANCER PATIENTS RECEIVING NEPHROTOXIC CHEMOTHERAPY

Abstract

Recurrence is a major clinical challenge in HNSCC; short-term relapse is related to poorer prognosis and few therapeutic choices. A recurrent radio genomic modeling framework to predict early recurrence in head and neck squamous cell carcinoma (HNSCC) patients is proposed in this paper. The proposed approach integrates the features extracted from medical imaging data, the features extracted from the genomic data, and the clinical data to reflect the tumor phenotype and underlying molecular behavior. The model aims to improve the prediction of the recurrence risk for each patient before or after treatment using imaging biomarkers, gene-expression profiles, mutation-related indicators and patient-level clinical features. The relationships between features in sequences and higher-dimensional feature spaces are learned using recurrent deep learning networks and contrasted with the best current state of the art in machine learning. The framework can aid precision oncology by defining patients who are at risk of early recurrence that may warrant increased monitoring and/or increased individual treatment planning or adjuvant therapy. To conclude, radio genomic modelling could be a useful tool for improving the level of accurate recurrence prediction and clinical decision making in the treatment of HNC.