Comparison of accuracy between augmented reality/mixed reality techniques and conventional techniques for epidural anesthesia using a practice phantom model kit.

BACKGROUND: This study used an epidural anesthesia practice kit (model) to evaluate the accuracy of epidural anesthesia using standard techniques (blind) and augmented/mixed reality technology and whether visualization using augmented/mixed reality technology would facilitate epidural anesthesia. METHODS: This study was conducted at the Yamagata University Hospital (Yamagata, Japan) between February and June 2022. Thirty medical students with no experience in epidural anesthesia were randomly divided into augmented reality (-), augmented reality (+), and semi-augmented reality groups, with 10 students in each group. Epidural anesthesia was performed using the paramedian approach with an epidural anesthesia practice kit. The augmented reality (-) group performed epidural anesthesia without HoloLens2Ⓡ and the augmented reality (+) group with HoloLens2Ⓡ. The semi-augmented reality group performed epidural anesthesia without HoloLens2Ⓡ after 30 s of image construction of the spine using HoloLens2Ⓡ. The epidural space puncture point distance between the ideal insertion needle and participant's insertion needle was compared. RESULTS: Four medical students in the augmented reality (-), zero in the augmented reality (+), and one in the semi-augmented reality groups failed to insert the needle into the epidural space. The epidural space puncture point distance for the augmented reality (-), augmented reality (+), and semi-augmented reality groups were 8.7 (5.7-14.3) mm, 3.5 (1.8-8.0) mm (P = 0.017), and 4.9 (3.2-5.9) mm (P = 0.027), respectively; a significant difference was observed between the two groups. CONCLUSIONS: Augmented/mixed reality technology has the potential to contribute significantly to the improvement of epidural anesthesia techniques.

Prediction of mortality risk of health checkup participants using machine learning-based models: the J-SHC study.

Early detection and treatment of diseases through health checkups are effective in improving life expectancy. In this study, we compared the predictive ability for 5-year mortality between two machine learning-based models (gradient boosting decision tree [XGBoost] and neural network) and a conventional logistic regression model in 116,749 health checkup participants. We built prediction models using a training dataset consisting of 85,361 participants in 2008 and evaluated the models using a test dataset consisting of 31,388 participants from 2009 to 2014. The predictive ability was evaluated by the values of the area under the receiver operating characteristic curve (AUC) in the test dataset. The AUC values were 0.811 for XGBoost, 0.774 for neural network, and 0.772 for logistic regression models, indicating that the predictive ability of XGBoost was the highest. The importance rating of each explanatory variable was evaluated using the SHapley Additive exPlanations (SHAP) values, which were similar among these models. This study showed that the machine learning-based model has a higher predictive ability than the conventional logistic regression model and may be useful for risk assessment and health guidance for health checkup participants.

Creation of an artificial intelligence model for intubation difficulty classification by deep learning (convolutional neural network) using face images: an observational study.

BACKGROUND: Tracheal intubation is the gold standard for securing the airway, and it is not uncommon to encounter intubation difficulties in intensive care units and emergency rooms. Currently, there is a need for an objective measure to assess intubation difficulties in emergency situations by physicians, residents, and paramedics who are unfamiliar with tracheal intubation. Artificial intelligence (AI) is currently used in medical imaging owing to advanced performance. We aimed to create an AI model to classify intubation difficulties from the patient's facial image using a convolutional neural network (CNN), which links the facial image with the actual difficulty of intubation. METHODS: Patients scheduled for surgery at Yamagata University Hospital between April and August 2020 were enrolled. Patients who underwent surgery with altered facial appearance, surgery with altered range of motion in the neck, or intubation performed by a physician with less than 3 years of anesthesia experience were excluded. Sixteen different facial images were obtained from the patients since the day after surgery. All images were judged as "Easy"/"Difficult" by an anesthesiologist, and an AI classification model was created using deep learning by linking the patient's facial image and the intubation difficulty. Receiver operating characteristic curves of actual intubation difficulty and AI model were developed, and sensitivity, specificity, and area under the curve (AUC) were calculated; median AUC was used as the result. Class activation heat maps were used to visualize how the AI model classifies intubation difficulties. RESULTS: The best AI model for classifying intubation difficulties from 16 different images was generated in the supine-side-closed mouth-base position. The accuracy was 80.5%; sensitivity, 81.8%; specificity, 83.3%; AUC, 0.864; and 95% confidence interval, [0.731-0.969], indicating that the class activation heat map was concentrated around the neck regardless of the background; the AI model recognized facial contours and identified intubation difficulties. CONCLUSION: This is the first study to apply deep learning (CNN) to classify intubation difficulties using an AI model. We could create an AI model with an AUC of 0.864. Our AI model may be useful for tracheal intubation performed by inexperienced medical staff in emergency situations or under general anesthesia.