Development of artificial intelligence-driven biosignal-sensitive cardiopulmonary resuscitation robot.

AIM OF THE STUDY: We evaluated whether an artificial intelligence (AI)-driven robot cardiopulmonary resuscitation (CPR) could improve hemodynamic parameters and clinical outcomes. METHODS: We developed an AI-driven CPR robot which utilizes an integrated feedback system with an AI model predicting carotid blood flow (CBF). Twelve pigs were assigned to the AI robot group (n = 6) and the LUCAS 3 group (n = 6). They underwent 6 min of CPR after 7 min of ventricular fibrillation. In the AI robot group, the robot explored for the optimal compression position, depth and rate during the first 270-second period, and continued CPR with the optimal setup during the next 90-second period and beyond. The primary outcome was CBF during the last 90-second period. The secondary outcomes were coronary perfusion pressure (CPP), end-tidal carbon dioxide level (ETCO2) and return of spontaneous circulation (ROSC). RESULTS: The AI model's prediction performance was excellent (Pearson correlation coefficient = 0.98). CBF did not differ between the two groups [estimate and standard error (SE), -23.210 ± 20.193, P = 0.250]. CPP, ETCO2 level and rate of ROSC also did not show difference [estimate and SE, -0.214 ± 7.245, P = 0.976 for CPP; estimate and SE, 1.745 ± 3.199, P = 0.585 for ETCO2; 5/6 (83.3%) vs. 4/6 (66.7%), P = 1.000 for ROSC). CONCLUSION: This study provides proof of concept that an AI-driven CPR robot in porcine cardiac arrest is feasible. Compared to a LUCAS 3, an AI-driven CPR robot produced comparable hemodynamic and clinical outcomes.

Impact of User's Background Knowledge and Polyp Characteristics in Colonoscopy with Computer-Aided Detection.

Background/Aims: We investigated how interactions between humans and computer-aided detection (CADe) systems are influenced by the user's experience and polyp characteristics. Methods: We developed a CADe system using YOLOv4, trained on 16,996 polyp images from 1,914 patients and 1,800 synthesized sessile serrated lesion (SSL) images. The performance of polyp detection with CADe assistance was evaluated using a computerized test module. Eighteen participants were grouped by colonoscopy experience (nurses, fellows, and experts). The value added by CADe based on the histopathology and detection difficulty of polyps were analyzed. Results: The area under the curve for CADe was 0.87 (95% confidence interval [CI], 0.83 to 0.91). CADe assistance increased overall polyp detection accuracy from 69.7% to 77.7% (odds ratio [OR], 1.88; 95% CI, 1.69 to 2.09). However, accuracy decreased when CADe inaccurately detected a polyp (OR, 0.72; 95% CI, 0.58 to 0.87). The impact of CADe assistance was most and least prominent in the nurses (OR, 1.97; 95% CI, 1.71 to 2.27) and the experts (OR, 1.42; 95% CI, 1.15 to 1.74), respectively. Participants demonstrated better sensitivity with CADe assistance, achieving 81.7% for adenomas and 92.4% for easy-to-detect polyps, surpassing the standalone CADe performance of 79.7% and 89.8%, respectively. For SSLs and difficult-to-detect polyps, participants' sensitivities with CADe assistance (66.5% and 71.5%, respectively) were below those of standalone CADe (81.1% and 74.4%). Compared to the other two groups (56.1% and 61.7%), the expert group showed sensitivity closest to that of standalone CADe in detecting SSLs (79.7% vs 81.1%, respectively). Conclusions: CADe assistance boosts polyp detection significantly, but its effectiveness depends on the user's experience, particularly for challenging lesions.

A remote-controlled automatic chest compression device capable of moving compression position during CPR: A pilot study in a mannequin and a swine model of cardiac arrest.

BACKGROUND: Recently, we developed a chest compression device that can move the chest compression position without interruption during CPR and be remotely controlled to minimize rescuer exposure to infectious diseases. The purpose of this study was to compare its performance with conventional mechanical CPR device in a mannequin and a swine model of cardiac arrest. MATERIALS AND METHODS: A prototype of a remote-controlled automatic chest compression device (ROSCER) that can change the chest compression position without interruption during CPR was developed, and its performance was compared with LUCAS 3 in a mannequin and a swine model of cardiac arrest. In a swine model of cardiac arrest, 16 male pigs were randomly assigned into the two groups, ROSCER CPR (n = 8) and LUCAS 3 CPR (n = 8), respectively. During 5 minutes of CPR, hemodynamic parameters including aortic pressure, right atrial pressure, coronary perfusion pressure, common carotid blood flow, and end-tidal carbon dioxide partial pressure were measured. RESULTS: In the compression performance test using a mannequin, compression depth, compression time, decompression time, and plateau time were almost equal between ROSCER and LUCAS 3. In a swine model of cardiac arrest, coronary perfusion pressure showed no difference between the two groups (p = 0.409). Systolic aortic pressure and carotid blood flow were higher in the LUCAS 3 group than in the ROSCER group during 5 minutes of CPR (p < 0.001, p = 0.008, respectively). End-tidal CO2 level of the ROSCER group was initially lower than that of the LUCAS 3 group, but was higher over time (p = 0.022). A Kaplan-Meier survival analysis for ROSC also showed no difference between the two groups (p = 0.46). CONCLUSION: The prototype of a remote-controlled automated chest compression device can move the chest compression position without interruption during CPR. In a mannequin and a swine model of cardiac arrest, the device showed no inferior performance to a conventional mechanical CPR device.

Density clustering-based automatic anatomical section recognition in colonoscopy video using deep learning.

Recognizing anatomical sections during colonoscopy is crucial for diagnosing colonic diseases and generating accurate reports. While recent studies have endeavored to identify anatomical regions of the colon using deep learning, the deformable anatomical characteristics of the colon pose challenges for establishing a reliable localization system. This study presents a system utilizing 100 colonoscopy videos, combining density clustering and deep learning. Cascaded CNN models are employed to estimate the appendix orifice (AO), flexures, and "outside of the body," sequentially. Subsequently, DBSCAN algorithm is applied to identify anatomical sections. Clustering-based analysis integrates clinical knowledge and context based on the anatomical section within the model. We address challenges posed by colonoscopy images through non-informative removal preprocessing. The image data is labeled by clinicians, and the system deduces section correspondence stochastically. The model categorizes the colon into three sections: right (cecum and ascending colon), middle (transverse colon), and left (descending colon, sigmoid colon, rectum). We estimated the appearance time of anatomical boundaries with an average error of 6.31 s for AO, 9.79 s for HF, 27.69 s for SF, and 3.26 s for outside of the body. The proposed method can facilitate future advancements towards AI-based automatic reporting, offering time-saving efficacy and standardization.