Related Experiment Video
Updated: Jun 28, 2026
09:31
High-Speed Ultraviolet Photoacoustic Microscopy for Histological Imaging with Virtual-Staining assisted by Deep Learning
Published on: April 28, 2022
3.5K
PSMA PET Evaluation with a Deep Learning Platform Compared with a Standard Image Viewer and Histopathology.
Daniel Koehler1,2, Farzad Shenas3, Markus Sauer3
1Department of Diagnostic and Interventional Radiology and Nuclear Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany; d.koehler@uke.de.
Summary
Artificial intelligence software aPROMISE showed high agreement for prostate-specific membrane antigen (PSMA) PET/CT interpretation, similar to standard viewers. This AI tool effectively segmented 92.1% of lesions, supporting its use in early prostate cancer evaluation.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Standardized prostate-specific membrane antigen (PSMA) PET/CT aids interpretation and reproducibility in prostate cancer management.
- Artificial intelligence (AI) offers potential to further enhance PSMA PET/CT analysis.
Purpose of the Study:
- To evaluate the performance of a deep learning software, aPROMISE, for PSMA PET/CT segmentation and reporting.
- To compare aPROMISE against a standard image viewer (IntelliSpace Portal [ISP]) in patients undergoing PSMA-radioguided surgery.
Main Methods:
- Retrospective analysis of [68Ga]Ga-PSMA-I&T PET/CT scans from 96 patients with biochemical recurrence after prostatectomy.
- Two readers assessed scans twice using ISP and twice using aPROMISE, calculating intra- and interrater agreement for miTNM stages.
- Correlation of imaging findings with histopathology as the standard of truth.
Main Results:
- High intrarater and interrater agreement rates (≥91.7% and ≥92.2%, respectively) were observed for both ISP and aPROMISE.
- The aPROMISE algorithm automatically segmented 129 out of 140 (92.1%) consensus-identified lesions.
- Substantial agreement (≥86.5%) was found between imaging and histopathology, with major staging differences in 34.4% of patients.
Conclusions:
- Intra- and interreader agreement for PSMA PET/CT evaluation were comparably high with both ISP and aPROMISE.
- AI-powered software like aPROMISE can effectively segment PSMA PET/CT lesions, demonstrating its utility as a supportive tool.
- AI applications show promise for enhancing PSMA PET/CT evaluation in early prostate cancer detection and management.
