The current study demonstrated that the examination of ADC histogram for the whole lesion based on FOCUS DWI could be assessed as a non-invasive tool for evaluating the extent of ALN involvement in early-stage breast cancer patients. The findings disclosed that four histogram parameters and the lesion size differed significantly among early-stage breast cancers at different ALN statuses. Furthermore, according to ROC curve analysis, the energy might be the most promising parameter for predicting ALN status in early-stage breast cancer among the significant parameters.
At present, the spin-echo single excitation (SS-EPI) sequence is commonly utilized for breast DWI acquisition. However, SS-EPI DWI images quality was not always satisfactory due to magnetic susceptibility artifacts . In breast imaging, image distortions and artifacts are evident due to anatomical complexity and isocentric scans . Our study obtained ADC maps for histogram analysis from FOCUS DWI sequences. FOCUS DWI is a new diffusion technique that uses two-dimensional space selective excitation pulses and 180˚ refocusing pulses . A previous study reported that in breast cancer, FOCUS DWI prominently improves image quality with a reduction of artifacts . In our study, the inter-observer variability study revealed that all parameters in the whole tumor histogram analysis had high inter-observer repeatability, with all parameters achieving ICCs greater than 0.900. However, a previous study  showed that skewness and kurtosis ICC scores were reasonably low (0.756 and 0.734, respectively) by demonstrating the efficacy of the whole-lesion technique based on SS-EPI DWI for discriminating Ki-67 expression in invasive breast cancer at T1 stage. To some extent, this phenomenon can be explained using FOCUS DWI technology; this enhances signal-to-noise ratio and spatial resolution of the picture, help overcome the partial volume effect, and thus make the whole tumor boundary delineation and semi-automatic segmentation results more accurate. Excellent inter-observer variability of whole-lesion histogram parameters depending on high-resolution FOCUS DWI technology is critical to ensuring the reliability of breast cancer quantification studies.
Breast cancers with different numbers of metastatic ALN require different surgical axillary treatment: patients with negative ALN metastasis do not require SLN biopsy or ALND. SLND is only for patients with 1–2 metastatic ALNs, and ALND is dedicated to patients with ≥ 3 metastatic ALNs [3, 26]. In our study, energy, maximum, 90 percentile, range, and lesion size revealed significant differences in predicting ALN status of early-stage breast cancer. Energy reflects the size of voxel value in the image. Prior research  showed that energy was significantly associated with histological grade and lymphovascular invasion of breast cancer. Zhao et al. proved that energy and total energy performed well in differentiating pN0 from pN1-2 nodal staging of the rectal cancer . In our study, the energy showed better diagnostic efficacy than the other parameters, which to some extent indicates that the energy value might be more linked to the malignant degree and invasiveness of cancers. A higher value of range reflects more variation of the intensity within VOI. In our study, the range values in N≥ 3 groups were significantly greater than those in N1 − 2 and N0 groups. Therefore, the range can reflect tumor heterogeneity to a certain extent.
According to a previous study, lower percentiles represent dense tumor cells, while higher percentiles reflect the areas of necrotic and edema components . In our study, the higher ADC percentiles (maximum, 90 percentile), corresponding to more necrotic and cystic components, showed closer correlations with ALN metastasis than lower ADC percentiles. Wang et al.  also identified that ADC 90 percentiles showed higher diagnostic efficacy for differentiating lymph node-positive and lymph node-negative groups of epithelial ovarian cancer. However, Liu et al. stated that lower ADC percentiles (such as 10 percentiles) showed more significant differences in gastric cancer patients with different N stages than higher percentiles . The above phenomena demonstrate that the probability of metastatic lymph nodes might be closely linked to different components of the primary tumor . The low ADC percentiles and high ADC percentiles played different roles in evaluating the prognosis of tumors in different parts. For breast cancer, tumors with more necrotic and cystic areas are more likely to have the aggressive biological behavior of ALN metastases.
Previous studies reported that tumor size was an independent prognostic factor of SLN . Tumor size was proportional to axillary lymph node metastasis, and each 0.1 cm higher in tumor size resulted in 4.29 times more likely to have SLN metastasis in breast cancer . Our study showed that the lesion size was the largest for the number of metastatic ALNs of > 3, followed by the number of metastatic ALNs of 1–2, and then by no metastasis, similar to previous studies. Therefore, rapid tumor growth accompanied higher malignancy, leading to ALN metastasis .
Numerous limitations are present in this investigation. First, the study is a retrospective analysis of data acquired from a prospective study, and there is inevitable patient selection bias. Second, the sample size for this study was rather small, and a greater sample size and multicenter data will be considered for ALN state evaluation in the future. Finally, we only used the traditional simple exponential model, which may lead to the bias of ADC values. In the future, we will attempt to add intravoxel incoherent motion, diffusion kurtosis imaging, and their obtained factors into our research.