This prospective study was approved by the Institutional Review Board (IRB) at Dacia Clinic (Oak Park, IL, USA). Bilateral axillary ultrasound was performed on 20 female subjects (2021–2024) using a linear probe (6-18 MHz) following the research protocol. A total of 85 ALNs were evaluated and 42 ALNs were used for the analysis. 43 ALNs were retrospectively excluded due to either incomplete data in 31 ALNs or errors in setting the axillary reference or ALN center in the remaining 12 ALNs. 

Axillary scanning was performed by 4 sonographers and 2 radiologists with experience in breast ultrasound, all ultrasound operators were right-handed. 2 of 6 operators scanned the same subject.  A 3D mapping device with magnetically tracked position sensors (BVN G-2000, MetriTrack Inc., USA) was integrated with the ultrasound machine. The device provided the automated 3D mapping of the ALN for the first operator, guided the second operator to re-identify the same ALN and measured the movement of the subjects and axillary references during scanning. Each ultrasound operator measured the 3 orthogonal diameters of each sampled lymph node, and the midpoint of the longest diameter, which was matched by both operators was assigned as the ALN 3D center. ALN mapping was conducted relative to the subject's body and reproducible axillary references (e.g. other axillary lymph nodes or axillary vascular structures, see Fig.1). The 3D probe position and orientation were matched between operators to identify and measure the position of ALNs (see Fig. 2).

The ALN 3D center displacement between the first and second ultrasound operator relative to the set axillary reference was used to evaluate the inter-operator positional reproducibility. The effect of displacement of  axillary references and body movement during scanning on the position reproducibility of ALNs was analyzed using statistical models (linear and logistic regression) and machine learning models (regression trees).