Real-Time, AI-Guided Photodynamic Laparoscopy Enhances Detection in a Rabbit Model of Peritoneal Cancer Metastasis

Abstract

Accurate diagnosis is essential for effective cancer treatment, particularly in peritoneal surface malignancies, where failure to detect metastatic lesions can mislead the treatment plan. This study assessed the diagnostic accuracy of staging laparoscopy using the integration of artificial intelligence (AI)-guided photodynamic diagnosis (PDD) with the photosensitizer Phonozen, activated at 405 nm in a rabbit model. To create peritoneal carcinomatosis, VX2 cells were inoculated laparoscopically into the peritoneum of female white New Zealand rabbits. Conventional and PDD-guided laparoscopy utilized a customized light source that emitted broad-spectrum white light or 405-nm blue light, respectively. The surgical procedure comprised a tripartite approach: exploration and labeling of suspected nodules under white-light visualization, identification of additional metastatic tumors under blue-excitation fluorescent light, and confirmatory open laparotomy to locate overlooked nodules by palpation. Our results showed that the initial experimental data from 371 nodules in 14 rabbits, comparing conventional diagnostic laparoscopy and PDD, showed increased detection sensitivity from 67% ± 1.9% (conventional) to 98% ± 0.7% (PDD) in the small-size nodule. In the second experimental data set from 265 nodules in 10 rabbits, the addition of a real-time AI algorithm further increased the sensitivity to 100% ± 0.0%. Combining PDD with AI enhances the detection of peritoneal cancer metastasis in staging laparoscopy.