Comprehensive Evaluation of Quantitative Measurements from Automated Deep Segmentations of PSMA PET/CT Images
Obed Korshie Dzikunu*, Amirhossein Toosi†, Shadab Ahamed†, Sara Harsini, François Bénard, Xiaoxiao Li, Arman Rahmim · 2025
Read Comprehensive Evaluation of Quantitative Measurements from Automated Deep Segmentations of PSMA PET/CT Images by Obed Korshie Dzikunu*, Amirhossein Toosi†, Shadab Ahamed†, Sara Harsini, François Bénard, Xiaoxiao Li, Arman Rahmim (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-16237.txt.
Description
Obed Korshie Dzikunu and co-authors (2025) check whether automatic tumor outlining on prostate cancer scans gives measurements doctors can trust, not just outlines that look right. Their data are 380 PSMA PET/CT scans from patients whose prostate cancer returned after treatment, judged by levels of a marker protein. Deep-learning software draws the lesions, and the study scores what comes out: peak and average uptake, total lesion activity, tumor volume, lesion count, and how far lesions have spread.
They train three network designs, U-Net, Attention U-Net, and SegResNet, each with four training objectives, including a weighted loss they propose. The Attention U-Net paired with that loss matched expert measurements most closely, with agreement scores of 0.90 to 0.99 on uptake measures. Model choice changes the numbers a clinic would act on, so segmentation tools should be graded on measurements, not overlap alone.
Catalog record
| CREATOR | Obed Korshie Dzikunu*, Amirhossein Toosi†, Shadab Ahamed†, Sara Harsini, François Bénard, Xiaoxiao Li, Arman Rahmim |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-16237