Acquired Visual Expertise Adopts and Enhances Face-Selective Cortex

Radiologists acquire expert-level performance in detecting abnormalities in chest X-rays over years of training. Holistic processing has been proposed to support this acquired visual expertise, but its neural implementation remains unclear. Because face perception is a canonical form of holistic/configural processing and reliably engages occipitotemporal regions such as the fusiform face area (FFA) and occipital face area (OFA), a key question is whether these regions also contribute to expertise-related holistic processing in a non-face professional domain. Here, we focused on face-selective regions as theoretically motivated regions of interest, while also using whole-brain searchlight decoding and resting-state functional connectivity analyses to characterize broader expertise-related changes across the brain. We combined behavioral and fMRI experiments in which radiologists and laypeople viewed faces, buildings, and chest X-rays under two manipulations of holistic processing: upright versus inverted presentation and amodally completed versus uncompleted images. Behavioral results indicated that radiologists adopted experience-dependent holistic strategies when interpreting chest X-rays. Multivariate pattern analysis (MVPA) revealed that disrupting holistic processing significantly reduced decoding accuracy for X-rays versus buildings in the left FFA and left OFA, indicating their involvement in expertise-related holistic processing. In addition, radiologists showed stronger face representations across stimulus forms in bilateral FFA. Resting-state analyses further revealed reorganization of intrinsic functional connectivity across the face-processing network. These findings suggest that radiological expertise engages left-lateralized holistic mechanisms in occipitotemporal cortex and is accompanied by stronger face representations and altered intrinsic connectivity within the face-processing network, supporting a model in which new expertise emerges through the adaptive reuse and amplification of pre-existing neural representations.