From gaze to decision dynamics: Mapping visual attention to diagnostic accuracy and confidence in dermatologic diagnoses

Understanding how visual attention supports diagnostic decision-making is critical in dermatology, a visually driven specialty. Although previous studies have examined gaze metrics, the dynamic relationship between gaze behavior, diagnostic accuracy, and confidence, particularly through latent attentional strategies, remains underexplored. Ten participants, comprising experts, residents, and medical students, diagnosed 50 different clinical images each, yielding 489 analyzable cases. Eye-tracking captured gaze behavior, which was analyzed using fixation measures, gaze sequence clustering, and hidden Markov models (HMMs) to uncover distinct attentional strategies. Longer and more frequent fixations on skin irregularities were negatively associated with diagnostic accuracy and confidence. Conversely, increased fixations on diagnostic choices indicated lower confidence. Two distinct gaze sequence clusters emerged: a consistent pattern linked to higher accuracy and confidence, and a more variable pattern associated with lower accuracy and confidence, despite similar expertise distributions. HMMs further revealed that the higher-performing cluster exhibited a structured progression of attention from examination to evaluation, whereas the lower-performing cluster showed prolonged focus on skin irregularities with fewer attentional shifts, indicating a less efficient visual strategy. This study advances our understanding of the relationship between visual attention and dermatological diagnosis. Effective diagnostic performance and metacognition were significantly associated with both the spatial and temporal allocation of attention. These findings offer insights into designing decision-support systems and enhancing dermatology training.