Automatic Surgical Skill Assessment System Based on Concordance of Standardized Surgical Field Development Using Artificial Intelligence

Surgeons vs ChatGPT: Assessment and Feedback Performance Based on Real Surgical Scenarios

INTRODUCTION

Artificial intelligence tools are being progressively integrated into medicine and surgical education. Large language models, such as ChatGPT, could provide relevant feedback aimed at improving surgical skills. The purpose of this study is to assess ChatGPT´s ability to provide feedback based on surgical scenarios.

METHODS

Surgical situations were transformed into texts using a neutral narrative. Texts were evaluated by ChatGPT 4.0 and 3 surgeons (A, B, C) after a brief instruction was delivered: identify errors and provide feedback accordingly. Surgical residents were provided with each of the situations and feedback obtained during the first stage, as written by each surgeon and ChatGPT, and were asked to assess the utility of feedback (FCUR) and its quality (FQ). As control measurement, an Education-Expert (EE) and a Clinical-Expert (CE) were asked to assess FCUR and FQ.

RESULTS

Regarding residents' evaluations, 96.43% of times, outputs provided by ChatGPT were considered useful, comparable to what surgeons' B and C obtained. Assessing FQ, ChatGPT and all surgeons received similar scores. Regarding EE's assessment, ChatGPT obtained a significantly higher FQ score when compared to surgeons A and B (p = 0.019; p = 0.033) with a median score of 8 vs. 7 and 7.5, respectively; and no difference respect surgeon C (score of 8; p = 0.2). Regarding CE´s assessment, surgeon B obtained the highest FQ score while ChatGPT received scores comparable to that of surgeons A and C. When participants were asked to identify the source of the feedback, residents, CE, and EE perceived ChatGPT's outputs as human-provided in 33.9%, 28.5%, and 14.3% of cases, respectively.

CONCLUSION

When given brief written surgical situations, ChatGPT was able to identify errors with a detection rate comparable to that of experienced surgeons and to generate feedback that was considered useful for skill improvement in a surgical context performing as well as surgical instructors across assessments made by general surgery residents, an experienced surgeon, and a nonsurgeon feedback expert.

Real-time detection of active bleeding in laparoscopic colectomy using artificial intelligence

BACKGROUND

Most intraoperative adverse events (iAEs) result from surgeons' errors, and bleeding is the majority of iAEs. Recognizing active bleeding timely is important to ensure safe surgery, and artificial intelligence (AI) has great potential for detecting active bleeding and providing real-time surgical support. This study aimed to develop a real-time AI model to detect active intraoperative bleeding.

METHODS

We extracted 27 surgical videos from a nationwide multi-institutional surgical video database in Japan and divided them at the patient level into three sets: training (n = 21), validation (n = 3), and testing (n = 3). We subsequently extracted the bleeding scenes and labeled distinctively active bleeding and blood pooling frame by frame. We used pre-trained YOLOv7_6w and developed a model to learn both active bleeding and blood pooling. The Average Precision at an Intersection over Union threshold of 0.5 (AP.50) for active bleeding and frames per second (FPS) were quantified. In addition, we conducted two 5-point Likert scales (5 = Excellent, 4 = Good, 3 = Fair, 2 = Poor, and 1 = Fail) questionnaires about sensitivity (the sensitivity score) and number of overdetection areas (the overdetection score) to investigate the surgeons' assessment.

RESULTS

We annotated 34,117 images of 254 bleeding events. The AP.50 for active bleeding in the developed model was 0.574 and the FPS was 48.5. Twenty surgeons answered two questionnaires, indicating a sensitivity score of 4.92 and an overdetection score of 4.62 for the model.

CONCLUSIONS

We developed an AI model to detect active bleeding, achieving real-time processing speed. Our AI model can be used to provide real-time surgical support.

Nationwide Association of Surgical Performance of Minimally Invasive Esophagectomy With Patient Outcomes

Importance

Suboptimal surgical performance is hypothesized to be associated with less favorable patient outcomes in minimally invasive esophagectomy (MIE). Establishing this association may lead to programs that promote better surgical performance of MIE and improve patient outcomes.

Objective

To investigate associations between surgical performance and postoperative outcomes after MIE.

Design, Setting, and Participants

In this nationwide cohort study of 15 Dutch hospitals that perform more than 20 MIEs per year, 7 masked expert MIE surgeons assessed surgical performance using videos and a previously developed and validated competency assessment tool (CAT). Each hospital submitted 2 representative videos of MIEs performed between November 4, 2021, and September 13, 2022. Patients registered in the Dutch Upper Gastrointestinal Cancer Audit between January 1, 2020, and December 31, 2021, were included to examine patient outcomes.

Exposure

Hospitals were divided into quartiles based on their MIE-CAT performance score. Outcomes were compared between highest (top 25%) and lowest (bottom 25%) performing quartiles. Transthoracic MIE with gastric tube reconstruction.

Main Outcome and Measure

The primary outcome was severe postoperative complications (Clavien-Dindo ≥3) within 30 days after surgery. Multilevel logistic regression, with clustering of patients within hospitals, was used to analyze associations between performance and outcomes.

Results

In total, 30 videos and 970 patients (mean [SD] age, 66.6 [9.1] years; 719 men [74.1%]) were included. The mean (SD) MIE-CAT score was 113.6 (5.5) in the highest performance quartile vs 94.1 (5.9) in the lowest. Severe postoperative complications occurred in 18.7% (41 of 219) of patients in the highest performance quartile vs 39.2% (40 of 102) in the lowest (risk ratio [RR], 0.50; 95% CI, 0.24-0.99). The highest vs the lowest performance quartile showed lower rates of conversions (1.8% vs 8.9%; RR, 0.21; 95% CI, 0.21-0.21), intraoperative complications (2.7% vs 7.8%; RR, 0.21; 95% CI, 0.04-0.94), and overall postoperative complications (46.1% vs 65.7%; RR, 0.54; 95% CI, 0.24-0.96). The R0 resection rate (96.8% vs 94.2%; RR, 1.03; 95% CI, 0.97-1.05) and lymph node yield (mean [SD], 38.9 [14.7] vs 26.2 [9.0]; RR, 3.20; 95% CI, 0.27-3.21) increased with oncologic-specific performance (eg, hiatus dissection, lymph node dissection). In addition, a high anastomotic phase score was associated with a lower anastomotic leakage rate (4.6% vs 17.7%; RR, 0.14; 95% CI, 0.06-0.31).

Conclusions and Relevance

These findings suggest that better surgical performance is associated with fewer perioperative complications for patients with esophageal cancer on a national level. If surgical performance of MIE can be improved with MIE-CAT implementation, substantially better patient outcomes may be achievable.

Artificial intelligence for automatic surgical phase recognition of laparoscopic gastrectomy in gastric cancer

PURPOSE

This study aimed to classify laparoscopic gastric cancer phases. We also aimed to develop a transformer-based artificial intelligence (AI) model for automatic surgical phase recognition and to evaluate the model's performance using laparoscopic gastric cancer surgical videos.

METHODS

One hundred patients who underwent laparoscopic surgery for gastric cancer were included in this study. All surgical videos were labeled and classified into eight phases (P0. Preparation. P1. Separate the greater gastric curvature. P2. Separate the distal stomach. P3. Separate lesser gastric curvature. P4. Dissect the superior margin of the pancreas. P5. Separation of the proximal stomach. P6. Digestive tract reconstruction. P7. End of operation). This study proposed an AI phase recognition model consisting of a convolutional neural network-based visual feature extractor and temporal relational transformer.

RESULTS

A visual and temporal relationship network was proposed to automatically perform accurate surgical phase prediction. The average time for all surgical videos in the video set was 9114 ± 2571 s. The longest phase was at P1 (3388 s). The final research accuracy, F1, recall, and precision were 90.128, 87.04, 87.04, and 87.32%, respectively. The phase with the highest recognition accuracy was P1, and that with the lowest accuracy was P2.

CONCLUSION

An AI model based on neural and transformer networks was developed in this study. This model can identify the phases of laparoscopic surgery for gastric cancer accurately. AI can be used as an analytical tool for gastric cancer surgical videos.

Automatic surgical phase recognition-based skill assessment in laparoscopic distal gastrectomy using multicenter videos

BACKGROUND

Gastric surgery involves numerous surgical phases; however, its steps can be clearly defined. Deep learning-based surgical phase recognition can promote stylization of gastric surgery with applications in automatic surgical skill assessment. This study aimed to develop a deep learning-based surgical phase-recognition model using multicenter videos of laparoscopic distal gastrectomy, and examine the feasibility of automatic surgical skill assessment using the developed model.

METHODS

Surgical videos from 20 hospitals were used. Laparoscopic distal gastrectomy was defined and annotated into nine phases and a deep learning-based image classification model was developed for phase recognition. We examined whether the developed model's output, including the number of frames in each phase and the adequacy of the surgical field development during the phase of supra-pancreatic lymphadenectomy, correlated with the manually assigned skill assessment score.

RESULTS

The overall accuracy of phase recognition was 88.8%. Regarding surgical skill assessment based on the number of frames during the phases of lymphadenectomy of the left greater curvature and reconstruction, the number of frames in the high-score group were significantly less than those in the low-score group (829 vs. 1,152, P < 0.01; 1,208 vs. 1,586, P = 0.01, respectively). The output score of the adequacy of the surgical field development, which is the developed model's output, was significantly higher in the high-score group than that in the low-score group (0.975 vs. 0.970, P = 0.04).

CONCLUSION

The developed model had high accuracy in phase-recognition tasks and has the potential for application in automatic surgical skill assessment systems.