Abstract 4161: Validation of the ViewsML vIHC platform for biomarker classification and intensity binning: A pilot collaboration with Histowiz’s PathologyMap platform

Key Points

• This research aims to validate ViewsML’s vIHC platform for predicting biomarker expression from H&E images.
• Collaboration with Histowiz’s PathologyMap platform for data integration.
• Analysis of 50 lung carcinoma cases using formalin-fixed paraffin embedded samples.
• Training machine learning models on paired H&E and IHC whole slide images to predict biomarker expression intensity.
• Evaluation of model performance using ROC-AUC on a test set.
• Model achieved high classification performance with ROC-AUC values of 0.94 for CD68, 0.92 for PanCK, and 0.95 for SMA.
• Predicted per-cell stain intensity across millions of cells, reflecting biological variability.
• Validated vIHC as a reliable alternative to traditional immunohistochemistry.