Liver Cancer Prediction Using Machine Learning: Enhancing Early Detection and Survival Analysis

Abstract

Liver cancer is still one of the most lethal cancers in the world, with consistently increasing rates in the United States that are caused by rising rates of obesity, rates of hepatitis infection, and liver disease that is associated with alcohol. Early detection of liver cancer is crucial for improving patient survival because liver cancer is typically found in advanced stages with dismal survival rates and few treatment choices. The overall objective of this study was to create and test machine-learning models for liver cancer diagnosis and survival prediction. The research focused on machine learning in the U.S. health system using patient data with different demographic and clinical backgrounds. The dataset for this study is a rich patient dataset collected with great care to support machine learning model development for liver cancer detection and survival prediction. It had detailed patient demographic data, including age, gender, ethnicity, and geographic origin, that are crucial for population-based risk factor identification and liver cancer disparities. Additionally, the dataset has large medical history records of pre-existing conditions of chronic infections with hepatitis B and C, cirrhosis, NAFLD, diabetes, and alcohol use disorder that are crucial liver cancer risk factors. Genetic factors like SNPs and gene expression patterns that are implicated in liver cancer are also present to study genetic susceptibility to disease development and progression. Clinical test results like ultrasounds, CT and MRI images, and biomarker levels like AFP and DCP form a robust platform for diagnostic and prediction modeling. The dataset is obtained from multiple high-quality sources like Electronic Health Records (EHRs) of top health centers, anonymized patient databases of hospitals, and national cancer databases like the Surveillance, Epidemiology, and End Results (SEER) Program. In addressing the dual objectives of liver cancer detection and survival prediction, a combination of machine learning models was employed, with each chosen for its specific strength. Accuracy, precision, recall, and F1-score were used for classification tasks to test whether liver cancer was identified by the models. XG-Boost performs better than both models with the highest accuracy and with strong precision, recall, and F1 scores, representing its strength in classification. The use of AI tools in the U.S. health system can revolutionize methods of early detection for liver cancer and address one of oncology's biggest challenges. With machine learning models that are trained on rich databases, clinicians can be equipped with potent diagnostic tools that enhance their ability to diagnose liver cancer in its earliest and most curable stages. The use of machine learning models in clinical decision support systems (CDSS) is a revolutionary opportunity to improve liver cancer treatment in the U.S. health system. The application of AI-based predictive models in liver cancer treatment has important public health and policy implications for the United States.