Leveraging Artificial Neural Networks and Support Vector Machines for Accurate Classification of Breast Tumors in Ultrasound Images

利用人工神经网络和支持向量机对超声图像中的乳腺肿瘤进行精确分类

阅读:1

Abstract

Background and Aim Breast cancer is a leading cause of cancer-related deaths among women, and ultrasound imaging is crucial for early detection. However, variability in interpretation can affect diagnosis. Therefore, this study compared the performance of artificial neural networks (ANNs) and support vector machines (SVMs) in classifying breast tumors using ultrasound images. Method This comparative study was conducted from June 1, 2023, to June 1, 2024, using a convenience sampling method. Data were gathered from the Center for Breast Diseases at Nanakali Hospital in Erbil, Kurdistan Region of Iraq, and a publicly available dataset from Kaggle. ANN and SVM models were then applied to classify the tumors. Statistical analysis was performed using R (R Foundation for Statistical Computing, Vienna, Austria) and IBM SPSS Statistics for Windows, Version 28.0 (Released 2021; IBM Corp., Armonk, New York, United States), with performance metrics such as accuracy, sensitivity, specificity, and Kappa coefficient calculated for both models. Results The ANN model achieved an accuracy of 87.78%, with a sensitivity of 86.67% and a specificity of 88.89%. The SVM model demonstrated an accuracy of 86.67%, with a higher specificity of 95.56% but a lower sensitivity of 77.78%. Both models showed substantial agreement between predicted and actual classifications, with Kappa coefficients of 75.56% for ANN and 73.33% for SVM. The mean, skewness, and area were identified as the most important variables for the ANN model, while solidity, circularity, and perimeter were the most critical features of the SVM model. Conclusions The results indicated that ANN had a marginally higher accuracy than SVM in classifying breast tumors. It is recommended to further optimize these models for clinical use, improve the integration of machine learning in medical imaging, and expand the dataset to enhance model generalizability and robustness.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。