The value of Demetics ultrasound-assisted diagnosis system in diagnosis of breast lesions and in assessment Ki-67 status of breast cancer.

PURPOSE: This study aims to explore the diagnostic efficiency of the Demetics for breast lesions and assessment of Ki-67 status. MATERIAL: This retrospective study included 291 patients. Three combined methods (method 1: upgraded BI-RADS when Demetics classified the breast lesion as malignant; method 2: downgraded BI-RADS when Demetics classified the breast lesion as benign; method 3: BI-RADS was upgraded or downgraded according to Demetrics' diagnosis) were used to compare the diagnostic efficiency of two radiologists with different seniority before and after using Demetics. The correlation between the visual heatmap by Demetics and the Ki-67 expression level of breast cancer was explored. RESULTS: The sensitivity, specificity, and area under curve (AUC) of diagnosis by Demetics, junior radiologist and senior radiologist were 89.5%, 83.1%, 0.863; 76.9%, 82.4%, 0.797 and 81.1%, 89.9%, 0.855, respectively. Method 1 was the best for senior radiologist, which increased AUC from 0.855 to 0.884. For junior radiologist, Method 3 was the best method, improving sensitivity (88.8% vs. 76.9%) and specificity (87.2% vs. 82.4%). Demetics paid more attention to the peripheral area of breast cancer with high expression of Ki-67. CONCLUSION: Demetics has shown good diagnostic efficiency in the assisted diagnosis of breast lesions and is expected to further distinguish Ki-67 status of breast cancer.

Deep Learning with Transformer or Convolutional Neural Network in the Assessment of Tumor-Infiltrating Lymphocytes (TILs) in Breast Cancer Based on US Images: A Dual-Center Retrospective Study.

This study aimed to explore the feasibility of using a deep-learning (DL) approach to predict TIL levels in breast cancer (BC) from ultrasound (US) images. A total of 494 breast cancer patients with pathologically confirmed invasive BC from two hospitals were retrospectively enrolled. Of these, 396 patients from hospital 1 were divided into the training cohort (n = 298) and internal validation (IV) cohort (n = 98). Patients from hospital 2 (n = 98) were in the external validation (EV) cohort. TIL levels were confirmed by pathological results. Five different DL models were trained for predicting TIL levels in BC using US images from the training cohort and validated on the IV and EV cohorts. The overall best-performing DL model, the attention-based DenseNet121, achieved an AUC of 0.873, an accuracy of 79.5%, a sensitivity of 90.7%, a specificity of 65.9%, and an F1 score of 0.830 in the EV cohort. In addition, the stratified analysis showed that the DL models had good discrimination performance of TIL levels in each of the molecular subgroups. The DL models based on US images of BC patients hold promise for non-invasively predicting TIL levels and helping with individualized treatment decision-making.