A Spatially Adaptive Neural Framework for Lossy X-Ray Compression with Preservation of Diagnostic Fidelity
| dc.rights.license | CC-BY-NC-ND | |
| dc.contributor.advisor | Santos Silva, W.J. dos | |
| dc.contributor.author | Wurth, Isabelle | |
| dc.date.accessioned | 2025-08-21T00:06:34Z | |
| dc.date.available | 2025-08-21T00:06:34Z | |
| dc.date.issued | 2025 | |
| dc.identifier.uri | https://studenttheses.uu.nl/handle/20.500.12932/49901 | |
| dc.description.sponsorship | Utrecht University | |
| dc.language.iso | EN | |
| dc.subject | This thesis presents a neural compression method for chest X-rays that preserves diagnostic content. By combining Spatial Feature Transform layers with quality maps from BioMedCLIP, the model allocates more bits to clinically relevant regions. Evaluation using a DenseNet121 classifier shows improved diagnostic consistency and image quality over traditional codecs, especially at low bitrates. | |
| dc.title | A Spatially Adaptive Neural Framework for Lossy X-Ray Compression with Preservation of Diagnostic Fidelity | |
| dc.type.content | Master Thesis | |
| dc.rights.accessrights | Open Access | |
| dc.subject.courseuu | Artificial Intelligence | |
| dc.thesis.id | 51995 |
