MITOS-ATYPIA-14
MITOS-ATYPIA-14: Automated Mitosis Detection & Nuclear Atypia Scoring
This project is based on the MITOS-ATYPIA-14 Grand Challenge, which focuses on developing artificial intelligence methods for automated detection of mitotic figures and grading of nuclear atypia in breast cancer histopathology images. These tasks are central components of tumor grading and directly impact clinical decision-making.
Mentor Details:
Dr. Chen Sagiv
Mentor Details:
Requirments:
- Develop AI models for accurate detection and localization of mitotic figures.
- Predict nuclear atypia grades from histopathology image patches.
- Evaluate model performance using standardized challenge metrics.
- Produce reproducible pipelines suitable for benchmarking and research use.
Problem Statement
Manual identification of mitotic figures and assessment of nuclear atypia in H&E-stained histopathology slides is time-consuming, subjective, and prone to inter-observer variability. As digital pathology adoption increases, scalable and reproducible automated solutions are needed to support pathologists and improve diagnostic consistency.
Project Objectives
- Develop AI models for accurate detection and localization of mitotic figures.
- Predict nuclear atypia grades from histopathology image patches.
- Evaluate model performance using standardized challenge metrics.
- Produce reproducible pipelines suitable for benchmarking and research use.
Technical Scope
- Preprocessing of high-resolution H&E-stained pathology images
- Patch extraction and color normalization
- Deep learning-based object detection and classification models
- Post-processing and scoring logic
- Model evaluation using precision, recall, F1-score, and grading accuracy
Required Knowledge and Prerequisites
Core Requirements:
- Python programming
- Machine learning fundamentals
- Computer vision and deep learning (PyTorch / TensorFlow)
Recommended Background:
- Digital pathology or medical imaging
- Object detection and segmentation architectures
- Experience with microscopy or histology data
Project Difficulty and Expected Level
Overall Difficulty: varies
This project is well-suited for:
Teams of 2–4 students
This project can also be done coding free with the DeePathology STUDIO.
Expected Outcomes
- A trained mitosis detection model
- Nuclear atypia grading predictions
- Benchmark evaluation results
- Reproducible codebase and documentation