Date: May 2024
Source: PhD Thesis. University of Houston Institutional Repository.
Abstract: The ability to utilize 3D imaging technology to retrieve and compare cases of previous patients can improve the clarity and precision of consultations. This dissertation introduces a novel system of 3D torso image retrieval (3DTIR), aimed at enhancing the consultation process for breast reconstruction surgeries. The developed method integrates suggestive contours, the structure tensor, and mesh simplification using MeshCNN for 3D model retrieval. Extensive experiments involving both proprietary (140 3dMD-captured patient scans) and public benchmark datasets (Shrec15, Faust) validated the effectiveness of the proposed method. Results were assessed by three reconstructive surgeons for medical appropriateness on a 4-point scale, with a mean score of 3.77 across all patients, and the proposed method statistically outperformed three existing state-of-the-art retrieval methods (p < 0.05) when evaluated on the 3dMD-captured dataset.
Article: 3D Torso Image Retrieval of Prior Patient Cases for Enhancing Consultation about Breast Reconstruction Surgery.
Authors: Nassima Noufail. Committee: Chair-Fatima A. Merchant; Members-Shishir K. Shah, Guoning Chen, Gregory P. Reece. Department of Computer Science, College of Natural Sciences and Mathematics, University of Houston.