Disease Dependencies - TobiasSchmidtDE/DeepL-MedicalImaging GitHub Wiki

Motivation: There is strong clinical evidence that labels exhibit dependencies.

In ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases it is shown that in their attempt to build a hospital-scale x-ray database they find this assumption to be true.

In Learning to diagnose from scratch by exploiting dependencies among labels the problem of multi-class learning with dependencies is treated as a sequence prediction with fixed length, hence they use LSTMs. The formulation of their LSTM is similar to those used in image and video captioning. See Architecture Implementations for more information.