Uncertainty Estimation in Semantic Segmentation of US Images

Uncertainty Estimation in Semantic Segmentation of US Images

Team: Bishesh Khanal, Lavsen Dahal and Aayush Kafle

2D US is the most common imaging modality for cardiovascular diseases and fetal scans. The portability and relatively low-cost nature of Ultrasound (US) enable the US devices to be made widely available. However, acquiring and interpreting US images is operator dependent, limiting its use to only places where experts are present. Deep Learning (DL) has shown some promise in aiding automated view classification, and structure and function assessment from US images.

Although these recent works show promise in developing computer-guided acquisition and automated interpretation of US images, most of these methods do not model and estimate uncertainty which can be important when testing on data coming from a distribution further away from that of the training data. Uncertainty estimates can be beneficial both during the image acquisition phase (by providing real-time feedback to the operator on acquired image’s quality), and during automated measurement and interpretation. The performance of uncertainty models and quantification metric may depend on the prediction task and the models being compared.

We are developing novel methods for gaining insight in uncertainty modelling for semantic image segmentation from US images.

Relevant Publications

  1. Dahal, L., Kafle, A., & Khanal, B. (2020). Uncertainty Estimation in Deep 2D Echocardiography Segmentation. arXiv preprint arXiv:2005.09349. pdf
Research Themes: Transforming Global health with AI (TOGAI)
Project Category: Computer Vision, Machine Learning, Medical Imaging