Machine utilization in diagnostic ultrasound for triple-negative breast cancer: a systematic review.
Code: G-1456
Authors: Yasamin Ahmadi * ℗, Amir Mohammad Chekeni, Zahra Darabi
Schedule: Not Scheduled!
Tag: Biomedical Signal Processing
Download: Download Poster
Abstract:
Abstract
Introduction: Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer that has limited treatment options due to the lack of estrogen receptors (ER), progesterone receptors (PR), and HER2, and its accurate identification is of great importance. Recent advances in machine learning (ML) and its application in medical diagnosis, especially in ultrasound image analysis, have provided new possibilities for non-invasive diagnosis of TNBC. This systematic review aimed to evaluate the application of machine learning in ultrasound image analysis to diagnose TNBC and differentiate it from other breast cancer subtypes. Methods: A review was conducted independently by two people based on PICO criteria and aligned with the research objective and based on the PRISMA checklist, using PubMed, CINAHL, Medline, Web of Science, Google Scholar search engine SID databases, and Boolean operators. The time limit between 2018 and 2024 was determined using the MESH keywords "triple negative breast cancer", "machine learning" and "diagnostic ultrasound". After reviewing the inclusion and exclusion criteria and critically assessing the quality of the selected articles, a total of 8 articles were included in the study. Results: The results showed that the combination of grayscale and color Doppler image features improved the performance of the triple negative breast cancer (TNBC) detection model, so that the accuracy of the model increased to 88%, the sensitivity reached 86.96% and the detection accuracy reached 82.91%. By adjusting the parameters, the XGBoost model was able to show the highest negative prediction value (NPV) of 98.1% and the lowest false negative result rate (1.9%). Also, the use of radiomics technique by extracting specific features from ultrasound images had a great impact on improving the performance of the models. Conclusion: The findings suggest that integrating machine learning with ultrasound data can be used as an effective tool for non-invasive and more accurate diagnosis of TNBC. Developing more comprehensive models and evaluating them in larger populations could help improve clinical decision-making and reduce the workload of physicians.
Keywords
Triple-negative Breast Cancer, Machine Learning
