A quantitative study of pathologists' perceptions towards artificial intelligence-assisted diagnostic system

The successful implementation of artificial intelligence-assisted diagnostic system (AIADS) in pathology relies not only on the maturity of AI technology but also on pathologists' cognition and acceptance of AI. However, research on pathologists' perceptions towards AIADS is limited. This study aims to explore pathologists' knowledge, attitudes, and practice toward AIADS and identify key factors influencing their willingness to use it, providing insights for the effective integration of AI technology in pathology. An online, nationwide, cross-sectional survey is to investigate pathologists' knowledge, attitudes and behavioral intention/practice regarding AIADS with a 5-point Likert scale. Descriptive analysis is used to present the results, while logistic regression examines factors influencing AIADS adoption. The mediating effect of attitude in the association between knowledge and behavioral intention is also explored. A total of 224 pathologists were surveyed, with 85 (37.9%) having used AIADS and 139 (62.1%) not using it. The mean scores for knowledge, attitude, and behavioral intention were 3.42 ± 0.97, 3.48 ± 0.44, and 3.47 ± 0.44, respectively. Pathologists who had used AIADS scored higher in knowledge, attitude, and behavioral intention, with clearer attitudes toward AIADS. Over 80% of pathologists supported the use of AIADS in clinical diagnostics, citing improved diagnostic speed and reduced workload as key reasons. The main concerns about AIADS were its diagnostic accuracy. Logistic regression analysis indicated that a greater likelihood of willingness to use AIADS was associated with not having used it before (OR=2.462, 95%CI 1.087-5.573), as well as with higher knowledge scores (OR=1.140, 95%CI 1.076-1.208) and more positive attitude scores (OR=1.119, 95%CI 1.053-1.189). Mediation analysis indicated an indirect path from knowledge to behavioral intention through attitude among individuals who have used AIADS, with the mediation effect accounting for 59.4%. In conclusion, most pathologists support the use of AIADS in clinical practice, but improvements in diagnostic performance are necessary. Enhancing pathologists' knowledge, attitudes, and user experience is crucial for the broader adoption of AIADS.