🤖 AI Summary
Nodules in breast ultrasound videos exhibit significant inter-slice appearance heterogeneity, making manual identification time-consuming and highly experience-dependent.
Method: We propose the first end-to-end nodule re-identification framework specifically designed for ultrasound video analysis. Our approach introduces a nodule-level dynamic feature extractor that jointly encodes local morphological characteristics and contextual information from surrounding glandular/ductal structures, coupled with unsupervised real-time spectral clustering to achieve cross-view matching of individual nodules.
Contribution/Results: This work pioneers the adaptation of person re-identification techniques to medical ultrasound, establishing a novel paradigm—“video feature extraction → nodule-level representation → online clustering-based grouping.” Evaluated on hundreds of clinical ultrasound videos, our method significantly improves nodule discrimination accuracy, enabling clinicians to rapidly localize and track the same nodule across multiple scanning planes, thereby providing a practical technical foundation for per-nodule precision analysis.
📝 Abstract
Ultrasound is a vital diagnostic technique in health screening, with the advantages of non-invasive, cost-effective, and radiation free, and therefore is widely applied in the diagnosis of nodules. However, it relies heavily on the expertise and clinical experience of the sonographer. In ultrasound images, a single nodule might present heterogeneous appearances in different cross-sectional views which makes it hard to perform per-nodule examination. Sonographers usually discriminate different nodules by examining the nodule features and the surrounding structures like gland and duct, which is cumbersome and time-consuming. To address this problem, we collected hundreds of breast ultrasound videos and built a nodule reidentification system that consists of two parts: an extractor based on the deep learning model that can extract feature vectors from the input video clips and a real-time clustering algorithm that automatically groups feature vectors by nodules. The system obtains satisfactory results and exhibits the capability to differentiate ultrasound videos. As far as we know, it's the first attempt to apply re-identification technique in the ultrasonic field.
