RECURRENCE RADIO GENOMIC MODELING FOR EARLY RECURRENCE PREDICTION IN HEAD AND NECK SQUAMOUS CELL CARCINOMA

Abstract

The acute kidney injury (AKI) is a common and serious event in intensive care units (ICUs) in cancer patients who are receiving nephrotoxic chemotherapy, which results in prolonged length of hospital stay, delay of treatment, elevated death rate, and costs to health care. It is a challenge to identify patients at risk early, due to the fact that renal deterioration may occur very rapidly and may not be appreciated until changes in blood urea or serum creatinine are clinically significant. This study presents a real-time machine learning framework that uses dynamic clinical, laboratory, medication and physiological data to predict AKI in patients with cancer in the ICU undergoing nephrotoxic chemotherapy. Predictive models were trained using patient-related parameters, including baseline renal function, serum creatinine, blood urea nitrogen, urine output, electrolyte concentrations, hemodynamic parameters, chemotherapy exposure, comorbidities and parameters related to monitoring in the intensive care unit. Several machine learning models were tested to find the best method to predict the onset of AKI. The aim of the proposed system, is to continuously assess risk, and to identify high risk patients before they become severely impaired. The results indicate that real-time machine learning could prove beneficial for early clinical decision making, better managing chemotherapy treatment, directing renal-protective treatment, and for enhancing patient prognosis within the context of critical oncology care. This is a proof-of-concept of the power of AI to improve individualized monitoring and prevent unnecessary kidney damage in vulnerable groups of cancer patients in the intensive care unit.