Allen Mo MD PhD

Using AI to Personalize Treatment for Liver Cancer
 

Allen Mo, MD, PhD, and Rafi Kabarriti, MD, are developing machine learning (ML) algorithms that will help clinicians with selecting optimal treatments for patients with hepatocellular carcinoma (HCC), the primary cause of liver cancer. Curative treatments for HCC are liver transplantation or surgery to remove the tumor. However, many patients with HCC who cannot tolerate transplantation or surgical resection undergo other less invasive treatment options, including radiofrequency ablation, transarterial chemoembolization, or stereotactic body radiotherapy (SBRT), an advanced radiation technique that has shown great promise for treating HCC.

At tumor board meetings, a multidisciplinary team of doctors, including surgeons, radiation oncologists, medical oncologists and interventional radiologists, considers the features and circumstances of each patient to determine the best course of treatment. For liver cancer, selecting the appropriate therapy can be challenging because patients typically have underlying poor liver function and thus pose unique challenges balancing treatment with differences in liver function, tumor size and location, and risk of treatment-related liver toxicity. Clinical guidelines provide a range of potential treatment options and offer limited guidance for specific recommendations. As a result, the recommendations from multidisciplinary discussion may often be subjective and inconsistent. Dr. Mo explains, “We want to help ensure that patients with HCC receive the best treatment at the right time by developing an artificial intelligence (AI) guided clinical support tool that relies on multiple objective measures to reproducibly aid doctors in complex treatment decision making.”

With support from ROI, Dr. Mo and Dr. Kabarriti are expanding on their previously published work and developing ML clinical support tools for two common clinical scenarios: patients with newly diagnosed HCC and patients who have received prior treatment for HCC and require additional therapy. To accomplish their goals, they are:

• Developing a ML tool that compares the risks and benefits of the most common HCC treatment options for a specific patient to predict progression free survival while considering toxicity to also assess liver function changes.
• Conducting a two-phase prospective observational trial that will compare the ML recommendations to the decisions made by the liver tumor board at their institution and determine if their ML tool can improve treatment recommendation consistency.

The team has extensive experience with treating patients for HCC, and their ML tools are based on a curated database that contains over 900 patients. They aim to use ML to make HCC treatment selection more consistent by better approximating tumor control and toxicity of different treatments. If their work is successful, patients could receive more personalized care with reduced treatment-related toxicity, and it may potentially lead to increased utilization of SBRT as a standard non-invasive treatment option for HCC.

Presentations
 

• Using Machine Learning-Based Boosting Algorithm to Predict Hepatotoxicity After Local Therapy and Association with Survival in Patients Undergoing Liver Directed Therapy was presented at the 2023 ACSO Annual Meeting.