A paper on “Contrastive Representations for Continual Learning of Fine-grained Histology Images” by the DART digital pathology and radiomics team has been presented at the Machine Learning in Medical Imaging (MLMI) pre-conference workshop at MICCAI 2021, the 24th International Conference on Medical Image Computing and Computer Assisted Intervention, September 27th to October 1st 2021. The annual MICCAI conference attracts world leading biomedical scientists, engineers, and clinicians from a wide range of disciplines associated with medical imaging and computer assisted intervention. See the Conference Proceedings here.

Paper Abstract: We show how a simple autoencoder based deep network with a contrastive loss can effectively learn representations in a continual/incremental manner with limited labelling. This is of particular interest to the biomedical imaging research community, for whom the visual task is often a binary decision (healthy vs. disease) with limited quantity data and costly labelling. For such applications, the proposed method provides a light-weight option of 1) representing patterns with relatively few training samples using a novel collaborative contrastive loss function 2) update the autoencoder based deep network in an unsupervised fashion for continual learning for new incoming data. We overcome the drawbacks of existing methods through planned technical design, and demonstrate the efficacy of the proposed method on three histology image classification tasks (lung, colon, breast cancer) with SOTA results.

Read the full paper here.

Lead author, Tapabrata (Rohan) Chakraborty says “It is great to present the first paper of the DART project at the MLMI workshop at MICCAI, a leading venue on engineering applications in biomedical imaging. This introductory methodological paper should motivate more translational articles from the DART project in the near future as we delve into clinical applications.”

Dr Chakraborty works on the DART work package on Digital Pathology AI and Radiomics Model Development.