Clinical Decision Support
Credit: FangXiaNuo/Getty Images

A clinical decision support system in oncology that combined the expertise of humans with AI-software that queries medical registries, has shown it is an effective aid for oncologists in their practice of medicine. Details of the system, called xDECIDE, were published today in the preview issue of AI in Precision Oncology, a new peer-reviewed journal from Mary Ann Liebert Inc.

xDECIDE is a clinical decision support system from oncology-focused AI platform company xCures. It’s “Human-AI Team” is designed to provide quick rankings of potential treatments for cancer patients via AI combined with human review to provide personalized treatment options for patients, while also employing outcomes tracking through an observational tracking protocol.

“The system identifies potentially effective treatment options individualized for each patient from the integration of real-world evidence, human expert knowledge and opinion, and scientific and clinical publications and databases, and offers a platform to learn from the experience of every patient,” the researchers wrote.

The system leverages the IRB-approved pan-cancer registry called XCELSIOR, in which patients have consented to provide their data, which provides data aggregation and continuous learning from patient outcomes and sharing of limited datasets for research. xDECIDE, leveraging natural language processing (NLP) and machine learning, aggregates and processing medical records to provide a structured care summary of standardized patient features. This information is then analyzed by an ensemble of models and produces a ranked list of treatment options for each patient. This output is then reviewed by human experts in the field—molecular pharmacologists and oncologists—acting as a virtual tumor board (VTB) to provide the final report of treatment options individualized to each patient, along with supporting rationales.

The team, comprising researchers from xCures, the Rare Cancer Research Foundation, Saint John’s Cancer Institute and Stanford University, note that the goal of offering a precision approach to cancer care has become increasingly complex in recent years as the number of genes and variants of significance that can inform treatment decisions increase 11% and 35%, respectively.

“Today, this equates to approximately 3,000 pages of new guideline information for a medical oncologist to absorb each year,” the researchers wrote. “When there are no approved treatment options, there may be dozens of relevant clinical trials for an oncologist and patient to chose from…The volume of actionable information, the speed of new information creation, and the need to quickly weigh the relative value of this information for a specific patient together demand computational support.” Further, they note, the rapid expansion of oncological knowledge is unlikely to abate.

The XCELSIOR pan-cancer registry and the xDECIDE engine were developed to provide just such support and the report released today details both the technologies used to quickly analyze data and the power of having a “human in the loop” to optimize the system’s performance.

“As the integration of real-world evidence, human expertise, and clinical databases evolve, platforms like xDECIDE will be central to improving patient outcomes and setting new standards in oncology care,” the authors concluded.

Also of Interest