Fig. 3: Performance comparison of BLDS, readers with and without the aid of BLDS in bone lesion detection and BM identification.

a Comparison between BLDS and six readers with different levels of expertise for lesion detection. b Difference analysis curve of BLDS, trainees and junior radiologists with and without the aid of BLDS by the evaluation of sensitivity and FPPC for lesion detection. c The weighted Jackknife alternative free-response receiver operating characteristic curves of trainees and junior radiologists without and with BLDS according to the level of readers experience in bone lesions detection. d Reading time of readers with and without assistance of BLDS in multicentre test datasets. e Comparison between BLDS and six readers with different levels of expertise for BM identification. f Difference analysis curve of BLDS, trainees and junior radiologists with and without the aid of BLDS by the evaluation of sensitivity and FPPC for BM identification. g Difference analysis curve of BLDS, trainees and junior radiologists with and without the aid of BLDS by the evaluation of patient-wise evaluation metrics for BM identification. h FROC curves of the pooled readers without and with BLDS in specific lesion detection. i Count of lesions in different groups based on bone lesion size and the sensitivity of readers with and without the BLDS in detecting bone lesions and BMs. j Count of patients in different groups based on bone lesion number per patient and the sensitivity of readers with and without the BLDS in detecting bone lesions and BMs. k Distribution of bone lesion size across different groups. l Distribution of the number of bone lesions across different groups. Source data are provided as a Source Data file (Source_data_Fig. 3.xlsx).