• computer vision refers to the analysis of images with AI models.

why is there few AI tools in the healthcare system

• image based prediction models have a high risk of bias.

  • lack of representative data
  • poorly labeled data

• overfit models perform well, but not on generalized data

how to increase the amount of data

• data augmentation

• transfer learning

• temporal data shifts cause drift in performance and requires continuous maintenances

Inflammatory bowel disease

• histological healing is being evaluated as secondary clinical endpoint

prevalence of disease is increasing in children

prediction of disease is the main objective of the research.

Pathology features allow for better prediction.

analysis pipeline

• color normalization

• Segmentation

• feature prediction and interpretable features.

• feature correlated with clinically relevant data

• two potential new genes

a machine learning approach to count and classify eosinophils.