Groundbreaking research by the UC Davis Health team has pioneered a Machine Learning model that accurately predicts liver cancer risk.
The collaboration between clinicians and data scientists at UC Davis Health has yielded a revolutionary Machine Learning model designed to predict the likelihood of hepatocellular carcinoma (HCC), a prevalent form of liver cancer.
The findings demonstrate how AI techniques can assist physicians in providing early risk HHC risk assessment for metabolic dysfunction-associated steatotic liver disease (MASLD) patients, helping to provide more personalised care.
Study co-author Aniket Alurwar, clinical informatics specialist at the UC Davis Center for Precision Medicine and Data Sciences, commented: “MASLD can lead to HCC, but the disease is quite sneaky, and it’s often unclear which patients face that risk.
“It doesn’t make sense to biopsy every patient with MASLD, but if we can segment for risk, we can track those people more closely and perhaps catch HCC early.”
Addressing the threat of MASLD
MASLD, formerly known as nonalcoholic fatty liver disease (NAFLD), presents a significant health challenge, particularly as it is intricately linked with metabolic disorders such as Type 2 diabetes.
With approximately a quarter of Americans affected by some form of MASLD, the stakes are high. The team’s study represents an exciting method of leveraging the capabilities of Machine Learning algorithms to identify disease risk.
Unveiling the power of predictive learning
Nine open-source algorithms were tested, and five were selected for additional evaluation and model development.
These chosen algorithms were trained using deidentified health data from 1,561 UC Davis Health MASLD patients, among whom 227 later developed HCC.
Subsequently, these top five algorithms underwent validation using data from 686 UC San Francisco patients sourced from deidentified medical records.
Among these patients, 176 were diagnosed with HCC. Ultimately, the Gradient Boosted Trees algorithm emerged as the most accurate prediction model, demonstrating superior statistical accuracy, sensitivity, and specificity.
Identifying new risk factors
While advanced liver fibrosis remains a prominent risk indicator for HCC, typified by elevated Fibrosis-4 Index (FIB-4) scores, the team’s analysis unearthed additional predictors, including high cholesterol, hypertension, bilirubin levels, and alkaline phosphatase (ALP) activity.
The team discovered various pathways leading to HCC, with high FIB-4 levels being the most evident. Interestingly, some patients with low FIB-4 levels but elevated cholesterol, bilirubin, and hypertension also developed HCC.
However, according to current guidelines, such patients would not typically receive preventive care measures.
This multifactorial approach significantly enhanced the predictive accuracy of the Machine Learning model to 92.23%.
With an impressive accuracy rate, the pilot model stands as a testament to the potential of AI in healthcare.
Notably, the model identified ‘low-risk’ MASLD patients who may still face heightened HCC susceptibility, challenging conventional screening protocols.
A roadmap to future AI advancements
Looking ahead, the UC Davis team remains committed to refining their model. By integrating clinical notes and exploring advanced AI techniques like natural language processing, they aim to further enhance predictive accuracy.
Ultimately, their vision extends to seamlessly integrating these advancements into electronic health records, empowering clinicians with real-time risk assessments.
