AI Model Predicts M2 Macrophage Levels and HCC Prognosis from Pathological Images
Author Information
Author(s): Tian Huiyuan, Tian Yongshao, Li Dujuan, Zhao Minfan, Luo Qiankun, Kong Lingfei, Qin Tao
Primary Institution: Henan Provincial People’s Hospital, Zhengzhou University People’s Hospital
Hypothesis
Can an artificial intelligence model predict M2 macrophage levels and the prognosis of hepatocellular carcinoma (HCC) using globally labeled pathological images?
Conclusion
The AI models effectively predicted M2 macrophage levels and HCC prognosis, suggesting a novel method for determining biomarker levels and forecasting prognosis without additional clinical tests.
Supporting Evidence
- The AI model achieved an AUC of 0.73 in predicting M2 macrophage levels.
- The predicted probabilities of M2 macrophage abundance were negatively associated with HCC prognosis.
- Using Lasso regression, the study identified significant clinical variables for prognosis prediction.
Takeaway
Researchers created a computer program that can look at images of liver cancer and guess how many special immune cells are present, which helps predict how well patients will do.
Methodology
The study used a weakly supervised AI model combining Masked Autoencoders with ResNet-32t to analyze Whole Slide Images for predicting M2 macrophage levels and HCC prognosis.
Potential Biases
Potential biases may arise from the variability in slide quality and the limited number of pathological images used.
Limitations
The study faced challenges with the consistency and quality of pathological slides and had a relatively small number of images, necessitating larger datasets for more robust results.
Participant Demographics
The study included 132 HCC patients from Henan Provincial People’s Hospital and 353 from the TCGA database, with a median age of 61 years.
Statistical Information
P-Value
p=0.031
Confidence Interval
95% CI: 0.59-0.87
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website