This course focuses on training medical image segmentation models with reduced labeled data. The learning outcomes include developing methods to train segmentation networks using mostly unlabeled data, achieving high accuracy with minimal labeled data, and deriving cardiac functional biomarkers from segmentations. The course teaches skills such as data augmentation, self-supervised learning, and pseudo-labeling. The teaching method involves a two-step process combining traditional supervision with self-supervised learning. The intended audience includes individuals interested in medical image analysis, machine learning in healthcare, and reducing the need for extensive labeling in training datasets.
MedAI - Training Medical Image Segmentation Models with Less Labeled Data - Sarah Hooper
Stanford University via YouTube
Overview
Syllabus
Intro
Many use cases for deep-learning based medical image segmentation
Goal: develop and validate methods to use mostly unlabeled data to train segmentation networks.
Overview Inputs: labeled data. S, and labeled data, Our approach two-step process using data augmentation with traditional supervision, self supervised learning and
Supervised loss: learn from the labeled data
Self-supervised loss: learn from the unlabeled data
Step 1: train initial segmentation network
Main evaluation questions
Tasks and evaluation metrics
Labeling reduction
Step 2: pseudo-label and retrain
Visualizations
Error modes
Biomarker evaluation
Generalization
Strengths
Taught by
Stanford MedAI
