-Original-

Differences in Exhibited Emotions between Junior Residents and Senior Doctors: Using an Artificial Intelligence-Based Imaging Analysis Tool

Naoto Matsuda^1,2, Sonoko Kirinoki^1,2 and Gen Takagi¹

¹Department of General Medicine and Health Science, Nippon Medical School, Tokyo, Japan
²Department of General Medicine, Dokkyo Medical University Saitama Medical Center, Saitama, Japan

Background: Students and junior residents often rely on subjective methods and evaluations to learn medical interviewing. Although facial expressions correlate with therapeutic outcomes, no study has systematically analyzed facial expressions in medical education. This study aimed to investigate the differences in facial expressions between senior physicians and junior residents during medical interviews using an artificial intelligence (AI)-based facial expression analysis tool.
Methods: Healthcare professionals at the Dokkyo Medical University Saitama Medical Center were recruited between November 2017 and October 2018. The medical interview duration was compared between junior and senior physicians. Facial emotions were analyzed using "Kokoro Sensor," an AI-based tool that classifies facial expressions into seven emotions—anger, contempt, disgust, fear, joy, sadness, and surprise—based on the proportion of frames classified as each emotion.
Result: Thirteen physicians participated, resulting in 20 video recordings from 8 junior residents and 19 from 5 senior physicians. The mean interview time was 18.95 ± 8.959 minutes for junior residents and 11.79 ± 8.073 minutes for senior physicians (p = 0.004). The percentage of time physicians' faces were recognized by the Kokoro Sensor (indicating when doctors looked at patients) was 9.4% for junior residents and 21.6% for senior physicians (p = 0.012). Emotional analysis showed a significant difference in the expression of surprise: 4.9% for junior residents and 15.9% for senior physicians (p = 0.011).
Conclusion: Facial emotion analysis revealed differences in facial expressions between junior residents and senior physicians during medical interviews. The findings suggest that an AI-based facial expression analysis tool can be used in education for both students and residents.

J Nippon Med Sch 2026; 93: 190-196

Keywords
facial emotion recognition, artificial intelligence, facial expression

Correspondence to
Naoto Matsuda
naoto-matsuda@nms.ac.jp

Received, October 17, 2025
Accepted, January 26, 2026