Artificial intelligence (AI) has shown promise in transforming health care, particularly in cardiology. However, there is a lack of high-quality evidence demonstrating its impact on crucial clinical outcomes. The purpose of this study was to synthesize existing evidence from randomized controlled trials (RCTs) on the application of AI in cardiology, evaluating its impact on key clinical outcomes. We conducted a systematic review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, searching MEDLINE, Web of Science, and the Cochrane Library from inception to November 2024. We included RCTs evaluating machine learning models compared to traditional methods in cardiovascular care. Primary outcomes focused on patient-important metrics, while secondary outcomes covered time and resource savings. Eleven RCTs met the inclusion criteria. Studies were conducted between 2021 and 2024, with 81.2% being multicenter trials. Five studies (45.5%) reported improvements in clinical events, 6 (54.5%) showed enhanced diagnostic accuracy and early detection, and 3 (27.3%) demonstrated improved resource utilization. This review highlights AI's potential to enhance cardiovascular care through improved early detection, diagnostic accuracy, and resource efficiency. However, the limited number of RCTs indicates a need for more high-quality studies to validate AI's effectiveness across various clinical domains.

Coronary artery disease (CAD) is a leading health challenge worldwide. Exercise stress testing is a foundational non-invasive diagnostic tool. Nonetheless, its variable accuracy prompts the exploration of more reliable methods. Recent advancements in machine learning (ML), including deep learning and natural language processing, have shown potential in refining the interpretation of stress testing data. Adhering to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, we conducted a systematic review of ML applications in stress electrocardiogram (ECG) and stress echocardiography for CAD prognosis. Medical Literature Analysis and Retrieval System Online, Web of Science, and the Cochrane Library were used as databases. We analysed the ML models, outcomes, and performance metrics. Overall, seven relevant studies were identified. Machine-learning applications in stress ECGs resulted in sensitivity and specificity improvements. Some models achieved rates of above 96% in both metrics and reduced false positives by up to 21%. In stress echocardiography, ML models demonstrated an increase in diagnostic precision. Some models achieved specificity and sensitivity rates of up to 92.7 and 84.4%, respectively. Natural language processing applications enabled the categorization of stress echocardiography reports, with accuracy rates nearing 98%. Limitations include a small, retrospective study pool and the exclusion of nuclear stress testing, due to its well-documented status. This review indicates the potential of artificial intelligence applications in refining CAD stress testing assessment. Further development for real-world use is warranted.

Objective: Mastoidectomy surgical training is challenging due to the complex nature of the anatomical structures involved. Traditional training methods based on direct patient care and cadaveric temporal bone training have practical shortcomings. 3D‐printed temporal bone models and augmented reality (AR) have emerged as promising solutions, particularly for mastoidectomy surgery, which demands an understanding of intricate anatomical structures. Evidence is needed to explore the potential of AR technology in addressing these training challenges.Methods: 21 medical students in their clinical clerkship were recruited for this prospective, randomized controlled trial assessing mastoidectomy skills. The participants were randomly assigned to the AR group, which received real‐time guidance during drilling on 3D‐printed temporal bone models, or to the control group, which received traditional training methods. Skills were assessed on a modified Welling scale and evaluated independently by two senior otologists.Results: The AR group outperformed the control group, with a mean overall drilling score of 19.5 out of 25, compared with the control group's score of 12 (p < 0.01). The AR group was significantly better at defining mastoidectomy margins (p < 0.01), exposing the antrum, preserving the lateral semicircular canal (p < 0.05), sharpening the sinodural angle (p < 0.01), exposing the tegmen and attic, preserving the ossicles (p < 0.01), and thinning and preserving the external auditory canal (p < 0.05).Conclusion: AR simulation in mastoidectomy, even in a single session, improved the proficiency of novice surgeons compared with traditional methods.Level of Evidence: NA Laryngoscope, 135:894–900, 2025

Capgras syndrome (CS) is a rare delusional misidentification syndrome in which individuals believe that another person has been replaced by an impostor. A retrospective chart review was conducted for all admissions to two male psychiatric inpatient departments in a tertiary hospital in Israel between August 1, 2024, and January 31, 2025. Cases with explicit documentation of CS were included. Demographic and clinical data were extracted and analyzed. Among 308 hospitalised patients, five (1.62%) met criteria for CS, with a mean age of 33.6 years. Three patients (60%) had schizophrenia spectrum disorders, and two (40%) had organic psychosis associated with left frontal brain lesions. Four patients (80%) presented during their first psychiatric admission after committing severe violence against family members perceived as impostors. Treatment response varied: CS resolved in some cases with antipsychotics alone, while others required Clozapine and adjunctive interventions. CS may be underrecognized but carries significant clinical implications, particularly due to its association with violence in first-episode psychosis. Early identification and tailored treatment are critical for risk reduction.

Background: Obesity is a heterogeneous chronic disease, yet it is still commonly defined and managed using body mass index alone. Identifying clinically meaningful subgroups may support more efficient, precise and practical care. Objective: To identify and characterize population-level heterogeneity among adults with obesity in a large Israeli healthcare cohort. Methods: In this cross-sectional study, we analyzed deidentified electronic health record data from the Leumit Obesity Registry. Adults with obesity who had at least one documented weight measurement and height or BMI record during 2024 were included. Demographic, socioeconomic, lifestyle, clinical, and treatment variables were analyzed using K-prototypes cluster analysis to identify subgroups within a mixed-data population. Results: The study included 68,203 adults with obesity. Five distinct patient clusters were identified: metabolically healthy young adults, 19% of the cohort; adults with low comorbidity, 18%; middle-aged adults with moderate comorbidity, 23%; multimorbid seniors with socioeconomic disadvantage, 24%; and higher socioeconomic status advanced age with high clinical burden, 16%. The clusters differed substantially in age, comorbidity burden, socioeconomic status, and obesity-treatment utilization. GLP-1 use was most common in the older multimorbid clusters, particularly among low SES. Dietitian use was lower in the oldest and sickest cluster, despite high disease burden. Bariatric surgery was overall relatively rare and was concentrated mainly in the younger clusters. Conclusions: Adults with obesity in this large population-based registry did not represent a single clinical group, but rather several distinct phenotypes with different clinical and treatment patterns. These findings support a shift toward more phenotype-informed obesity care and resource planning.