AI-Augmented Clinical Decision Support in a Patient-Centric Precision Oncology Registry

Autor: Mark Shapiro, Timothy J. Stuhlmiller, Bryan Federowicz, William Hoos, Asher Wasserman, Glenn Kramer, Zach Kaufman, Don Chuyka, Julie C. Friedland, Bill Mahoney, Al Musella, Mika Newton, Zachary Osking, J. M. Tenenbaum, Kenny K. Wong, Santosh Kesari, Jeff Shrager
Rok vydání: 2022
Popis: PurposexDECIDE is a clinical decision support system, accessed through a web portal and powered by a “Human-AI Team”, that offers oncology healthcare providers a set of treatment options personalized for their cancer patients, and provides outcomes tracking through an observational research protocol. This article describes the xDECIDE process and the AI-assisted technologies that ingest semi-structured electronic medical records to identify and then standardize clinico-genomic features, generate a structured personal health record (PHR), and produce ranked treatment options based on clinical evidence, expert insights, and the real world evidence generated within the system itself.MethodPatients may directly enroll in the IRB-approved pan-cancer XCELSIOR registry (NCT03793088). Patient consent permits data aggregation, continuous learning from clinical outcomes, and sharing of limited datasets within the research team. Assisted by numerous AI-based technologies, the xDECIDE team aggregates and processes patients’ electronic medical records, and applies multiple levels of natural language processing (NLP) and machine learning to generate a structured case summary and a standardized list of patient features. Next a ranked list of treatment options is created by an ensemble of AI-based models, called xCORE. The output of xCORE is reviewed by molecular pharmacologists and expert oncologists in a virtual tumor board (VTB). Finally a report is produced that includes a ranked list of treatment options and supporting scientific and medical rationales. Treating physicians can use an interactive portal to view all aspects of these data and associated reports, and to continuously monitor their patients’ information. The xDECIDE system, including xCORE, is self-improving; feedback improves aspects of the process through machine learning, knowledge ingestion, and outcomes-directed process improvement.ResultsAt the time of writing, over 2,000 patients have enrolled in XCELSIOR, including over 650 with CNS cancers, over 300 with pancreatic cancer, and over 100 each with ovarian, colorectal, and breast cancers. Over 150 VTBs of CNS cancer patients and ∼100 VTBs of pancreatic cancer patients have been performed. In the course of these discussions, ∼450 therapeutic options have been discussed and over 2,000 consensus rationales have been delivered. Further, over 500 treatment rationale statements (“rules”) have been encoded to improve algorithm decision making between similar therapeutics or regimens in the context of individual patient features. We have recently deployed the xCORE AI-based treatment ranking algorithm for validation in real-world patient populations.ConclusionClinical decision support through xDECIDE is available for oncologists to utilize in their standard practice of medicine by enrolling a patient in the XCELSIOR trial and accessing xDECIDE through its web portal. This system can help to identify potentially effective treatment options individualized for each patient, based on sophisticated integration of real world evidence, human expert knowledge and opinion, and scientific and clinical publications and databases.
Databáze: OpenAIRE