Automatic slice selection and diagnosis of breast ultrasound image using deep learning
Journal
Biomedical Signal Processing and Control
Journal Volume
97
Start Page
論文號碼 106688
ISSN
1746-8094
Date Issued
2024-11
Author(s)
Abstract
Breast cancer is the second most common cause of death in women worldwide; it can be postponed or even cured if detected, diagnosed, and treated in the early stage. Conventional car-based ultrasound (CBUS) is beneficial in clinical screening to assess the benignity and malignancy of breast cancers, which remains one tool for breast cancer assessment today. When investigating a suspicious region in the breast, operators meticulously record the abnormalities in the scanning process and suggest a biopsy examination for a highly suspicious lesion. However, a detailed screening records many slices, which might increase the review burden for clinicians. We proposed a computer-aided automatic slice selection and diagnosis system to relieve the burden on the medical staff. It consisted of two stages: the first stage, a Transformer-based model, was used to select suspicious slices from each patient's slice sequence. The second stage employed a modified convolutional neural network (CNN) model to assess whether suspicious slices were malignant or non-malignant (including normal or benign). In this study, we used 807 patients to evaluate our proposed methods, which contained variant numbers of slice sequences (range: 5 to 117 slices, average: 44 slices) and indicated interested images (representative slice). The experiment results show that the selection accuracy based on top-1 and top-5 scores was 74.35% and 97.27%, respectively, and demonstrated that the accuracy, sensitivity, specificity, and AUC of diagnosis were 79.85%, 80.13%, 79.80%, and 0.8641, respectively. This study proposes a CADx system with an automatic slice selection method to reduce physicians' review burden and recognize breast tumor malignancy.
SDGs
Publisher
Elsevier BV
Type
journal article