Diagnostic Accuracy for Breathing Sounds in Using an AI-Assisted Stethoscope Versus a Conventional Stethoscope

Purpose: This study compares the accuracy of dental students in diagnosing different breathing sounds using traditional versus artificial intelligence-assisted precordial stethoscopes (AIPS). Methods: The study involved 45 3rd- and 4th-year dental students who completed a pre-test to assess their baseline knowledge of breathing sounds. After watching a learning video, participants were randomized into 3 groups: conventional stethoscopes (Group A), AIPS (Group B) with audio, and AIPS with audio and video (Group C). Participants were evaluated in their diagnosis ability by listening to 10 recorded episodes of specific breathing sounds and then took post-tests to evaluate their learning. A post-survey was administered to evaluate the confidence and satisfaction of the participants with diagnostic responses, and to collect demographic information. Statistical significance was evaluated using a 1-way analysis of variance and paired sample t-test (P<0.05). Results: Significant improvements were observed in diagnostic accuracy (P<0.001) and knowledge (P=0.047) across all groups. AI groups demonstrated higher diagnostic accuracy for specific sounds (normal breathing, partial airway obstruction, and apnea). The confidence and satisfaction of the participants were strongly correlated with diagnostic accuracy (P<0.001), with those in Groups B and C reporting higher satisfaction. Group C responded fastest for apnea (P=0.021) and Group A for normal breathing (P=0.005), while Group A had the highest missing diagnosis proportion (53.3%). Conclusions: Artificial intelligence-assisted precordial stethoscopes enhance diagnostic accuracy and learning outcomes, particularly for students with lower baseline knowledge. These findings suggest that integrating artificial intelligence tools can improve the detection of respiratory complications during sedation, enhancing patient safety.