Automated Segmentation of Peritoneal Cancer Index Areas Using Deep Learning from CT Scans
Overview
This study introduces a deep learning method for automated segmentation of peritoneal cancer index (PCI) areas from CT scans.
Background
Peritoneal metastases (PM) are associated with poor prognosis and are primarily diagnosed through diagnostic laparoscopy. Accurate assessment of PM extent is crucial for treatment eligibility and monitoring response to therapies. Current imaging evaluations often underestimate disease burden.
Data Highlights
The study utilized a dataset of 62 contrast-enhanced CT scans from patients diagnosed with gastric, ovarian, or colorectal cancer. Each scan was annotated by clinical researchers to create segmentation masks for analysis.
Key Findings
The study presents a deep learning method for segmenting rPCI regions.
Convolutional and transformer-based architectures were compared for performance in segmentation tasks.
An anatomically constrained pipeline was developed to encode geometric boundary definitions of rPCI.
Performance was evaluated against interobserver agreement.
Clinical Implications
The automated segmentation method may facilitate more standardized evaluations of PM.
Conclusion
The introduction of deep learning for automated segmentation of PCI areas represents a significant advancement in the assessment of peritoneal metastases.
by Pieter C. Gort, Lotte J. S. Fleurkens-Ewals, Lenah D. Kampmeijer, Anna F. van Herwijnen, Marion W. Tops-Welten, Cris H. B. Claessens, Joost Nederend, Ignace H. J. T. De Hingh, Max J. Lahaye, Misha D. P. Luyer, Fons van der Sommen
Pancreatic adenocarcinoma, the most common form of pancreatic cancer, accounts for roughly 90 to 95 percent of cases. It's notoriously difficult to catch early, because symptoms often don't appear until the disease has already spread
