An Enhanced Residual Architecture for Automated Detection of Bone Metastasis on Whole-Body SPECT Images

Abstract

Single-photon emission computed tomography (SPECT) is a vital technology for diagnosing bone metastases and other skeletal diseases, playing a critical role in the early diagnosis and treatment of bone-related conditions. To address the challenges of physicians dedicating considerable time and effort to analyzing SPECT images, as well as the overly complex preprocessing steps in existing automatic diagnosis systems, an enhanced residual model that is better suited for clinical applications was proposed in this work. By incorporating a hybrid convolutional attention module and a pyramid pooling module into the backbone, the proposed model achieves comparable performance to the state-of-the-art methods while substantially simplifying the architecture. The experimental results show that the model achieved accuracy, recall, proportion of true positive and F-1 score of 0.8721, 0.7951, 0.9349 and 0.8594 respectively on our self-built dataset, MetaLung-SPECT. The results of the comparative experiments conducted on the public BS-80K dataset further corroborate that our proposed model achieves performance comparable to state-of-the-art methods.