Achyut Saroj, Founder, Consultant, and Author at AwareOnc, KOL Engagement and Medical Affairs Liaison at Tatva Health, shared a post on LinkedIn about a recent article by Jan Clusmann et al, published in Cancer Discovery:
“Can machine learning close the gap in liver cancer screening?
Hepatocellular carcinoma (HCC) is the third leading cause of cancer-associated death globally.
The tragedy?
It is often diagnosed too late for curative treatment because current screening focuses almost exclusively on patients already known to have liver cirrhosis.
However, the data tells a different story. In a large-scale study, nearly 70% of HCC cases were not diagnosed with chronic liver disease before their cancer diagnosis.
This highlights a massive gap in early detection that traditional protocols fail to address.
To address this, the authors developed PRE-Screen-HCC, an interpretable machine learning (ML) framework for global risk stratification that relies solely on routine clinical data.
Why is this a game-changer?
- Routine Data Power: By analyzing standard blood work (AST, ALT, GGT, platelets), demographics, and electronic health records, this model provides accurate risk assessment.
- Proven at Scale: Validated on over 900,000 individuals across the UKBio bank and the All of Us Research Program, the model significantly outperformed existing state-of-the-art risk scores.
- Transparent and Actionable: Unlike ‘black box’ AI, this framework is interpretable, showing clinicians exactly which factors. like age, waist circumference, or specific lab values are driving the risk score.
- Global Generalizability: The model maintained its high performance across diverse ethnic subgroups, making it a viable tool for under-resourced regions where early detection is most critical.
The authors believe that moving from a “cirrhosis-only” screening mindset to a data-driven, personalized risk model can transform HCC prognosis worldwide.
For fellow researchers and clinicians, the authors have released the full code and an interactive web calculator for external validation.”
Title: Machine Learning Predicts Hepatocellular Carcinoma Risk from Routine Clinical Data: A Large Population-Based Multicentric Study
Authors: Jan Clusmann, Paul-Henry Koop, David Zhang, Felix van Haag, Omar El Nahhas, Tobias Seibel, Laura Žigutytė, Apichat Kaewdech, Julien Calderaro, Frank Tacke, Tom Luedde, Daniel Truhn, Tony Bruns, Kai Schneider, Jakob Kather, Carolin Schneider
Read the Full Article on Cancer Discovery.
More posts featuring Achyut Saroj.