
Background Augmented Reality (AR) during Robotic-Assisted Radical Prostatectomy (RARP) provides surgical guidance by overlaying patient-specific 3D Virtual Models (VM) onto the intraoperative view. A main drawback is instrument occlusion, where VM overlaps the surgical tools, disturbing the augmented view. Deep learning (DL)-based desocclusion addresses this issue, but current implementation rely on expensive, specialized hardware. Methods We developed an AR desocclusion pipeline running on consumer-grade hardware (∼1150 US Dollars). For model selection, we trained and benchmarked 21 encoder architectures paired with a U-Net decoder. The selected model was integrated into a macOS application using CoreML inference on the Apple Neural Engine. Memory usage, thermal state and framerate were monitored on a continuous 81-minute console session. Photon-to-photon latency (from surgical field to the surgeon eye) was measured on the DaVinci® Xi using a microcontroller-phototransistor setup. Results FastViT-S12/Unet offered the highest accuracy among models meeting our inference-time budget (∼2 ms inference), achieving a Dice score of 95.1% (95% CI, 94.3-95.7) internally, 91.6% (95% CI, 90.2-92.9) and 90.2% (95% CI, 84.7-94.4) on externals tests, with a mean inference time of 1.89 ± 0.36 ms. The pipeline maintained 57.98 frames per second using 1.47 GB of unified memory with a nominal thermal state. On the surgical console, photon-to-photon median latency increased from 60 ms (Q1-Q3: 57.5-62.5) to 120.5 ms (Q1-Q3: 111-131) with the AR pipeline. Conclusions Real-time AR-RARP with DL-based desocclusion is achievable on low-cost hardware with an added latency remaining within ranges reported as technically acceptable for intraoperative use.
Aim Anastomotic leak remains a life-threatening complication after colorectal surgery. Static preoperative models show only modest discrimination, whereas clinical signs often appear late. We aimed to develop and internally validate dynamic machine-learning models integrating preoperative factors with serial postoperative biomarker and physiological trajectories. Method In this retrospective single-centre cohort, XGBoost models were developed using information available preoperatively and through POD1, POD3, or POD5. For the POD3 and POD5 models, patients with AL diagnosed by the respective time point were excluded so that each model estimated the risk of a subsequent leak. Internal validation used repeated stratified 5-fold cross-validation with 5 repeats, with preprocessing and threshold selection performed within the training data. Results Among 409 patients, 31 (7.6%) developed AL. AUROC was 0.696 ± 0.085 preoperatively and 0.714 ± 0.083 at POD1. At POD3, 405 patients remained at risk and 27 subsequently developed AL; AUROC was 0.781 ± 0.086. At POD5, 398 patients remained at risk and 20 subsequently developed AL; AUROC was 0.908 ± 0.047 and AUPRC 0.460 ± 0.152, with sensitivity 0.850 ± 0.177, specificity 0.792 ± 0.074, PPV 0.194 ± 0.081, and NPV 0.990 ± 0.011. Conclusion Serial postoperative trajectories improved internal risk stratification for AL, particularly by POD5. The high POD5 NPV should be interpreted in the context of low event prevalence and internal validation; external validation is required before the model is used to support clinical decisions.
BackgroundMicroscalpel based upon Airplasma® technology showed promising results in dermatology, plastic surgery and veterinary in terms of tissue lateral thermal damage (LTI). In this preclinical work, we aimed to explore its effect on central nervous system tissues which has not been established yet.MethodsSamples of glioma and meningioma tissue were cut by plasma microscalpel (PM) and bipolar forceps (BC). A pathologist, blinded to the assignment of samples, analyzed the presence of histological damage, that was the primary endpoint. This was a composite of 4 pathological aspects. A comparison between different types of cutting was performed.ResultsForty-three patients were included in the final analysis. Histological damage was observed in 14.29% of PM samples and in 46.34% of BC ones (P 0.001). A sensitivity analysis of the primary endpoint, based on significant damage in each pathological aspect, revealed significantly better results for PM samples.ConclusionsMicroscalpel based on Airplasma® technology showed less cellular damage if it is compared to BC. Even if it cannot supplant the bipolar cautery, this instrument could be important in microneurosurgery in terms of precise cutting, minimized temperature, very low lateral LTI and the possibility to reset mechanical forces.
Introduction Counting surgical instruments during surgical procedures is critical for ensuring patient safety. Our goal is to develop a machine learning-based computer vision algorithm to create a reliable, automated surgical instrument counting support system. We hypothesize that our machine learning algorithm, based on the YOLO (You Only Look Once) model, can effectively perform instrument counting in surgery videos, providing real-time feedback to the surgical team. Methods YOLO is a real-time object detection algorithm built on a convolutional neural network (CNN). To train the model, we record 100 minutes of open surgery video in simulation, including 600 instrument transfer events. Additionally, we collected 500 images of 6 different surgical instruments and obtained another 500 images from online surgeries. All images were annotated by a human expert to mark the name and locations of instruments. 80% data was allocated for model training and 20% for validation. Trained YOLO model was tested in a new surgical video for counting surgical instruments. Results The model achieved a satisfactory accuracy rate for instrument detection and counting. In a validation session, the model demonstrated a 72.7% success rate. In addition, the model classified large instruments more successfully than smaller instruments, such as small hemostats and needles, within the defined ROI. Conclusion YOLOv11-based computer vision model can accurately detect and count surgical instruments within a predefined region of interest in a simulated open surgical environment. Future research should focus on improving detection of small instruments, validating the system in real operating room environments.
Introduction Real-time artificial intelligence (AI) decision-support systems are increasingly proposed for use in endoscopic and minimally invasive surgery although how best to present such information intraoperatively is unclear. This study explored surgeon perspectives regarding the user-interface, interpretation and workflow integration of a prototype AI system (CLASSICA-OR) for real-time endoscopic cancer characterization during surgery. Methods Surgeons specialising in minimally invasive gastrointestinal surgery were asked for their first impressions in their evaluation of the CLASSICA-OR system within a high-fidelity surgical simulation environment. For this, participants viewed prerecorded colonoscopic imagery of significant rectal neoplasia while the AI analysis was applied in real-time to the video. Commentary and responses to semi-structured interview were recorded audiovisually for transcription and analysis using qualitative thematic analysis supported by NVivo (Lumivero). Results Eighteen surgeons participated, generating rich qualitative data. Four major themes emerged: (1) AI was viewed as a valuable but evolving decision-support tool; (2) institutional culture, governance and commercial realities were recognised as important determinants of adoption; (3) real-time perfusion-based lesion characterisation (foundational to CLASSICA-OR) was considered promising but with specific challenges; and (4) clarity, usability, transparency and workflow compatibility are surgeon priorities in decision-support displays. Participants emphasised minimised cognitive load, preserved surgeon autonomy, clear communication of uncertainty and seamless workflow integration. Preferences differed between simplified binary outputs and more detailed numerical displays supporting the need for adaptable interfaces. Conclusions Surgeons perceive real-time intraoperative AI decision support as promising but requiring thoughtful user-centred design. They articulate practical development and implementation insights for intraoperative AI systems.