Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form
D-Scholarship @ Pitt Banner and Links to Homepage

Trustworthy deep learning on medical images.

Hao, Degan (2025) Trustworthy deep learning on medical images. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img]
Preview
 PDF (Final Draft)
Primary Text
Download (2MB) | Preview

Abstract

Deep learning has achieved substantial advancements across a variety of medical imaging tasks, yet its performance can degrade when models are applied to new clinical settings or data distributions. For instance, a model trained exclusively on clean data might misclassify normal/benign cases as cancerous in the presence of adversarial perturbations. In another example, a model trained with data from one population may underperform on data from another population. This dissertation addresses the need for robust and trustworthy deep learning models in healthcare by developing and evaluating methods to improve model reliability in diverse medical imaging contexts. Key challenges include dealing with limited and potentially inaccurate labels, enhancing model interpretability, integrating clinical knowledge for improved lesion localization and survival prediction, and strengthening model robustness against adversarial attacks, spurious correlations, and data distribution shifts. By tackling these issues, the work contributes to the development of trustworthy deep learning models in medical applications.

Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Hao, Degandeh95@pitt.edudeh950000-0002-4001-2474
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairWu, Shandongwus3@upmc.edushw83
Committee MemberLinguraru, Marius Georgemlingura@childrensnational.org
Committee MemberZhou, LemingLeming.Zhou@pitt.edu
Committee MemberTafti, Ahmad P.tafti.ahmad@pitt.edu
Date: 7 January 2025
Date Type: Publication
Defense Date: 20 November 2024
Approval Date: 7 January 2025
Submission Date: 13 December 2024
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 111
Institution: University of Pittsburgh
Schools and Programs: School of Computing and Information > Intelligent Systems Program
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Medical imaging, artificial intelligence, deep learning, trustworthy, safety, explainability, bias,
Date Deposited: 07 Jan 2025 19:38
Last Modified: 07 Jan 2025 19:38
URI: http://d-scholarship.pitt.edu/id/eprint/47272

Metrics

Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item