Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Health Technology and Informatics | en_US |
| dc.contributor.advisor | Cai, Jing (HTI) | en_US |
| dc.contributor.advisor | Ren, Ge (HTI) | en_US |
| dc.creator | Zhu, Jiarui | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/14450 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | A clinically interpretable feature extraction approach based on 3D-gaussian for medical imaging analysis | en_US |
| dcterms.abstract | Although advanced deep learning models especially foundation models have dominated in most area of medical imaging analysis with remarkable performance, the implicitness of the feature representing and downstream generating process potentially bottlenecks their performance and draws serious clinical concerns on their interpretability during clinical deployment. In order to break the performance barrier and achieve cogent interpretability demonstration, in this study we develop a 3D gaussian representing framework with explicit knowledge-content-analyzable representations and knowledge-source-traceable generating process for medical imaging. Specifically, tailored frameworks are developed based on a newly emerging 3D scene representing technique for addressing long-standing challenges in the area of medical imaging segmentation, image-to-image translation and multimodal fusion. By leveraging the unique advantages that 3D gaussian representations possess, including highly-refined 3D representation, editable feature space, cross-domain adaptive distribution and a traceable generating process, these frameworks are able to address main challenges including incomplete or incorrect information extraction, suboptimal optimization target, ineffective training guidance and unreliable generated results. Exclusive experiments are conducted to evaluate the clinical efficacy of the proposed method with comparison to state-of-the-art methods, and the quantitative, qualitative and ablation analysis have demonstrated the superiority of the proposed method. This study provides a novel and robust interpretable paradigm for medical imaging analysis which is profoundly meaningful for a wide range of clinical applications. | en_US |
| dcterms.extent | xv, 124 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2026 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.LCSH | Diagnostic imaging -- Data processing | en_US |
| dcterms.LCSH | Image processing -- Digital techniques | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14450