Prototypical multiple instance learning for predicting lymph node metastasis of breast cancer from whole-slide pathological images.

Computerized identification of lymph node metastasis of breast cancer (BCLNM) from whole-slide pathological images (WSIs) can largely benefit therapy decision and prognosis analysis. Besides the general challenges of computational pathology, like extra-high resolution, very expensive fine-grained annotation, etc., two particular difficulties with this task lie in (1) modeling the significant inter-tumoral heterogeneity in BCLNM pathological images, and (2) identifying micro-metastases, i.e., metastasized tumors with tiny foci. Towards this end, this paper presents a novel weakly supervised method, termed as Prototypical Multiple Instance Learning (PMIL), to learn to predict BCLNM from WSIs with slide-level class labels only. PMIL introduces the well-established vocabulary-based multiple instance learning (MIL) paradigm into computational pathology, which is characterized by utilizing the so-called prototypes to model pathological data and construct WSI features. PMIL mainly consists of two innovatively designed modules, i.e., the prototype discovery module which acquires prototypes from training data by unsupervised clustering, and the prototype-based slide embedding module which builds WSI features by matching constitutive patches against the prototypes. Relative to existing MIL methods for WSI classification, PMIL has two substantial merits: (1) being more explicit and interpretable in modeling the inter-tumoral heterogeneity in BCLNM pathological images, and (2) being more effective in identifying micro-metastases. Evaluation is conducted on two datasets, i.e., the public Camelyon16 dataset and the Zbraln dataset created by ourselves. PMIL achieves an AUC of 88.2% on Camelyon16 and 98.4% on Zbraln (at 40x magnification factor), which consistently outperforms other compared methods. Comprehensive analysis will also be carried out to further reveal the effectiveness and merits of the proposed method.

Three-dimensional visual technique based on CT lymphography data combined with methylene blue in endoscopic sentinel lymph node biopsy for breast cancer.

BACKGROUND: The combined application of blue dye and radioisotopes is currently the primary mapping technique used for sentinel lymph node biopsy (SLNB) in breast cancer patients. However, radiocolloid techniques have not been widely adopted, especially in developing countries, given the strict restrictions on radioactive materials. Consequently, we carried out a retrospective study to evaluate the feasibility and accuracy of three-dimensional visualization technique (3DVT) based on computed tomography-lymphography (CT-LG) in endoscopic sentinel lymph node biopsy (ESLNB) for breast cancer. METHODS: From September 2018 to June 2020, 389 patients who underwent surgical treatment of breast cancer in our department were included in this study. The CT-LG data of these patients were reconstructed into digital 3D models and imported into Smart Vision Works V1.0 to locate the sentinel lymph node (SLN) and for visual simulation surgery. ESLNB and endoscopic axillary lymph node dissection were carried out based on this new technique; the accuracy and clinical value of 3DVT in ESLNB were analyzed. RESULTS: The reconstructed 3D models clearly displayed all the structures of breast and axilla, which favors the intraoperative detection of SLNs. The identification rate of biopsied SLNs was 100% (389/389). The accuracy, sensitivity, and false-negative rate were 93.83% (365/389), 93.43% (128/137), and 6.57% (9/137), respectively. Upper limb lymphedema occurred in one patient 3 months after surgery during the 12-month follow-up period. CONCLUSIONS: Our 3DVT based on CT-LG data combined with methylene blue in ESLNB ensures a high identification rate of SLNs with low false-negative rates. It, therefore, has the potential to serve as a new method for SLN biopsy in breast cancer cases.

Imaging features of sentinel lymph node mapped by multidetector-row computed tomography lymphography in predicting axillary lymph node metastasis.

INTRODUCTION: Accurately assessing axillary lymph node (ALN) status in breast cancer is vital for clinical decision making and prognosis. The purpose of this study was to evaluate the predictive value of sentinel lymph node (SLN) mapped by multidetector-row computed tomography lymphography (MDCT-LG) for ALN metastasis in breast cancer patients. METHODS: 112 patients with breast cancer who underwent preoperative MDCT-LG examination were included in the study. Long-axis diameter, short-axis diameter, ratio of long-/short-axis and cortical thickness were measured. Logistic regression analysis was performed to evaluate independent predictors associated with ALN metastasis. The prediction of ALN metastasis was determined with related variables of SLN using receiver operating characteristic (ROC) curve analysis. RESULTS: Among the 112 cases, 35 (30.8%) cases had ALN metastasis. The cortical thickness in metastatic ALN group was significantly thicker than that in non-metastatic ALN group (4.0 ± 1.2 mm vs. 2.4 ± 0.7 mm, P < 0.001). Multi-logistic regression analysis indicated that cortical thickness of > 3.3 mm (OR 24.53, 95% CI 6.58-91.48, P < 0.001) had higher risk for ALN metastasis. The best sensitivity, specificity, negative predictive value(NPV) and AUC of MDCT-LG for ALN metastasis prediction based on the single variable of cortical thickness were 76.2%, 88.5%, 90.2% and 0.872 (95% CI 0.773-0.939, P < 0.001), respectively. CONCLUSION: ALN status can be predicted using the imaging features of SLN which was mapped on MDCT-LG in breast cancer patients. Besides, it may be helpful to select true negative lymph nodes in patients with early breast cancer, and SLN biopsy can be avoided in clinically and radiographically negative axilla.