An Analysis of Multitask Deep Learning Models for Histopathology

Artificial Intelligence has the potential to streamline and facilitate numerous processes in the medical field, increasing the quality of life for millions and potentially saving lives. One area which requires a lot of effort when it comes to diagnosing severe diseases is Histopathology. Recently, a modern learning strategy, Multitask learning, was applied in Digital Histopathology in order to obtain relevant medical information from histological images. This paradigm is able to increase performance by learning multiple objectives simultaneously resulting in more general features. The resulting methods reduce overfitting and computational complexity while increasing data efficiency making it a suitable choice for the high-dimensionality, low sample size sets from the medical field. The aim of this work is to present novel multitask approaches with applications in Histopathology, analyse them, showcase their advantages and drawbacks and identify possible future research directions.