An AI-based system was developed to assess patient-submitted postoperative wound images, aiming to streamline the detection of surgical site infections (SSI). Among 6,060 patients studied, the model achieved an impressive incision detection accuracy of 94% and 73% accuracy for SSI detection. This technology demonstrated robust performance in image quality assessment and showed comparable efficacy across […]
Category: Digital Surgery and Telemedicine
AI model accurately assesses surgical skills in laparoscopic cholecystectomy
An AI-based surgical phase recognition model achieved 82.3% accuracy in assessing the skill levels of surgeons performing laparoscopic cholecystectomy. Novice surgeons demonstrated significantly longer dissection times compared to experts, with similar patterns observed in clipping and cutting tasks. The model proved effective in categorizing surgical proficiency, identifying key metrics such as phase duration and confidence […]
AI Enhances Prognostic Assessments in Resected Pancreatic Cancer
AI-powered spatial analysis of tumor-infiltrating lymphocytes (TILs) in resected pancreatic ductal adenocarcinoma (PDAC) shows promise for prognostic assessment. Among 304 patients, those with an immune-inflamed phenotype demonstrated significantly longer overall and recurrence-free survival compared to other immune profiles. High intratumoral TIL density correlated with extended overall survival (median 52.47 months). The study suggests that AI […]
New deep learning models excel in organ donation image segmentation
Deep learning models, detectron2 and yolov8, significantly improved automated segmentation in organ donation photography, achieving internal validation Intersection over Union (IoU) scores of 0.93 and 0.94 for kidneys, and 0.97 for livers. Compared to traditional methods, which peaked at IoU 0.54, these new models demonstrated superiority in accuracy and speed, completing tasks in just 0.13-1.5 […]
Deep learning model enhances 3D segmentation of pancreatic cancer
A newly developed deep learning auto-segmentation model effectively segments pancreatic cancer and surrounding structures to assist surgical planning. In a study involving 275 patients, the model achieved a Dice similarity coefficient of 75.4 in internal and 75.6 in external validation. High segmentation accuracy was noted in the pancreas parenchyma but lower in pancreatic cancer lesions. […]
AI-enhanced coaching improves laparoscopic surgery training
A new study introduces SmartCoach, an AI-assisted program designed to enhance surgical coaching for laparoscopic pancreatoduodenectomy. By automating video analysis and facilitating real-time assessments, this approach aims to overcome barriers such as limited expert feedback and the labor-intensive nature of traditional coaching. Preliminary data highlighted surgeons’ limited awareness of coaching principles and the advantages of […]
AI Enhances Real-Time Detection of Nerve Traction During Surgery
A novel AI system demonstrated effectiveness in real-time detection of excessive traction on the recurrent laryngeal nerve during robotic-assisted esophagectomy. In a pilot study, the AI accurately identified potential nerve damage in 84.4% of surgical scenes, providing crucial feedback earlier than conventional monitoring methods. This proactive approach could facilitate timely adjustments in surgical technique, helping […]
AI Enhances Early Detection of Pancreatic Cancer
Findings indicate that AI-driven imaging models significantly improve the detection of pancreatic ductal adenocarcinoma (PDA) through advanced radiomics and deep learning techniques. These models can identify subtle pre-diagnostic signs on CT scans months to years before clinical symptoms arise, addressing current limitations in sensitivity and interobserver variability. Although AI presents a promising avenue for earlier […]
AI-assisted system improves biliopancreatic disease detection
A novel AI-assisted endoscopic ultrasound (EUS) system demonstrated superior performance in recognizing biliopancreatic stations and segmenting anatomical structures. Testing with 45,737 EUS images revealed the mean teacher algorithm achieved 95.60% accuracy in internal tests and 92.72% in external validation for station recognition. Additionally, U-Net v2 emerged as the optimal model for segmentation. Notably, the system […]
Artificial intelligence enhances preoperative assessment of thyroid conditions
Advances in artificial intelligence are revolutionizing the preoperative evaluation of thyroid nodules, significantly improving diagnostic accuracy. By enhancing imaging, cytopathology diagnostics, and prognostic assessments, AI minimizes reliance on clinical judgment and reduces diagnostic errors. The ability to process large volumes of data enables a comprehensive analysis of thyroid conditions, facilitating the development of tailored treatment […]