Liver Metastases Disease Detection Based on MobileNetV2 and Swin Transformer Using SVM Classifier

Abstract

Liver metastasis is a significant challenge in oncology that needs accurate and prompt detection for effective treatment. Developing computer vision, machine learning, and deep learning techniques contributes to early detection of liver metastases. This paper presents a method for detecting liver metastases in three stages: feature extraction, feature fusion, and classification. Firstly, MobileNetV2, an efficient and lightweight convolutional neural network, has been used as a feature extraction technique due to its ability to handle high-resolution medical images in low computational complexity. Also, the Swin transformer has been applied for feature extraction due to its ability to have contextual information and long-range dependencies to improve feature extraction. Secondly, feature fusion was applied between extracted features to get complementary information from both features. Finally, the support vector machine (SVM) was applied during the classification stage. This method was trained and tested on private datasets and achieved 98.41% accuracy.