Office of Technology Transfer – University of Michigan

Algorithm to Identify Patients with Hepatocellular Carcinoma

Technology #5396
  1. Non-proprietary disclosure (NPD/marketing abstract) Assessing Patient Risk for Hepatocellular Carcinoma Approximately 10,000 people are diagnosed with hepatocellular carcinoma (HCC) every year. The prognosis is usually poor because the carcinoma can be completely removed in only 25% of cases. Screening for HCC is typically performed by closely monitoring patients' who are at elevated risk for cirrhosis of the liver, either due alcohol abuse, HCV, HBC, HIV etc. This process, however, is neither a good predictor of the presence of HCC in patients nor is it an effective allocation of hospital resources.

Predictive Modeling for Assessing Patient Risk for HCC Researchers in the Department of Statistics and Doctors at the University of Michigan Hospital had developed a machine learning algorithm that predicts the presence of HCC in patients. The predictive model uses results from standard laboratory tests as predictors of HCC. In a clinical trial from 2004 to 2010, the team demonstrated that the algorithm outperforms the standard methods that employ COX regression analysis.

Applications: - Identification of patients with HCC

Advantages: - More accurate that the current regression methods to predict the presence of HCC - Easily incorporated into electronic medical record systems