We propose L-MAE, a longitudinal masked auto-encoder that learns temporal representations from sequential retinal image pairs to predict diabetic retinopathy progression.
A multi-modal deep learning framework combining OCT and fundus imaging for AMD progression prediction, evaluated on the MARIO MICCAI 2024 challenge dataset.
amdmulti-modaloctprogression-prediction
LatiM: Longitudinal Representation Learning in Continuous-Time Models for Medical Image Analysis
LatiM is a continuous-time longitudinal model for learning representations from sequential medical images, enabling disease progression prediction from irregular time-series.
longitudinalcontinuous-timemiccaimedical-imaging
Longitudinal Self-Supervised Learning using Neural Ordinary Differential Equation
Self-supervised longitudinal learning via Neural ODEs for retinal image sequences applied to diabetic retinopathy follow-up and progression prediction.
self-supervisedneural-odelongitudinalmiccai
LMT: Longitudinal Mixing Training for Disease Progression Prediction using Multi-Source Data
Patent covering a deep learning method for longitudinal analysis of medical images, enabling progression monitoring and prediction for degenerative diseases.