*4.3. Limitations*

Due to the retrospective set-up of this study not all clinical, imaging, and pathology data were available for all patients. We had access to all the FDG PET scans, but the scans of other imaging modalities were not always present. In those cases, the available report of the radiologist was used to compare lesions on the different imaging modalities. Furthermore, of the 267 lesions investigated, 155/267 lesions were pathologically verified (reference method), whereas for 112/267 lesions, only additional imaging and/or follow-up data were available, which precludes a definitive diagnosis regarding these lesions. However, our separate analyses for both groups yielded results in a similar range, supporting the chosen approach. For the visual analysis, we also included all axillary lymph nodes, benign or malignant, as verified according to the pathology report. However, in the case of multiple avid lymph nodes on the FDG PET scan, it can be difficult to match the exact lymph node that was pathologically proven benign or malignant to the correct lesion on the scan. In that case, it was assumed that the lesion that is most avid on the scan is also most likely the one of which biopsy or cytology is performed, and this lesion could also be quantified. Of the remaining lesions, only the number of affected lymph nodes were taken into account in the analysis.
