Responsibilities include:

• Established the state-of-the-art performance of semi-supervised domain adaptation (SSDA) and unsupervised domain adaptation (UDA) methods for generalizability of algorithms for video-based assessment (VBA) of intraoperative skill in a critical step in cataract surgery.
• Evaluated the utility of UDA with maximum mean discrepancy (MMD) for prediction tasks that use temporal models.

Responsibilities include:

• Applied Semi-Supervised VoxelMorph framework on the CT scans of ET to do automated registration and segmentation.
• Build a completed pipeline to evaluate the performance of deep learning models from the clinical views.
• Developed an automated pipeline to build Statistical Shape Models (SSMs) from a pool of CT scans of patients to facilitate quantitative analysis and visual exploration of anatomical variations of eustachian tube (ET).
• Constructed anatomical ET shapes using Point Distribution Model, and captured principal modes of shape variability by Principal Component Analysis (PCA).

Responsibilities include:

• Developed algorithms for generalizability of skill assessment algorithms on cataract surgery study.
• De-identified and processed cataract surgical videos.