Timezone: »

Unsupervised Approaches for Out-Of-Distribution Dermoscopic Lesion Detection
Max Torop · Sandesh Ghimire · Dana H Brooks · Octavia Camps · Milind Rajadhyaksha · Kivanc Kose · Jennifer Dy

There are limited works showing the efficacy of unsupervised Out-of-Distribution (OOD) methods on complex medical data. Here, we present preliminary findings of our unsupervised OOD detection algorithm, SimCLR-LOF, as well as a recent state of the art approach (SSD), applied on medical images. SimCLR-LOF learns semantically meaningful features using SimCLR and uses LOF for scoring if a test sample is OOD. We evaluated on the multi-source International Skin Imaging Collaboration (ISIC) 2019 dataset, and show results that are competitive with SSD as well as with recent supervised approaches applied on the same data.

Author Information

Max Torop (Northeastern University)
Sandesh Ghimire (Northeastern University)
Dana H Brooks (Northeastern University)

Electrical and Computer Engineering, Northeastern University, co-founder of the Signal Processing, Imaging, Reasoning, and Learning (SPIRAL) group there, and co-Director of the Simulation and Estimation Core of the Center for Integrative Biomedical Computing headquartered at University of Utah. His research includes application of signal and image processing and machine learning to medical and biological imaging, modeling, inverse problems,and optimization.

Octavia Camps (Northeastern University)
Milind Rajadhyaksha (Memorial Sloan Kettering Cancer Center)
Kivanc Kose (Memorial Sloan Kettering Cancer Center)
Jennifer Dy (Northeastern University)

More from the Same Authors