Comparative Study on the Efficiency of Using LB-FCN and Contrastive Learning for Detecting Bone Tumor in Bone Scans

Nowadays, healthcare improvement has a big impact on the business sector through the reduction of healthcare costs and the creation of opportunities for companies to develop new technology for the medical equipment analysis of scintigraphy images. This technological improvement currently has a huge impact on biomedical science, whereas a lot of concern has shifted to detecting bone metastasis disease. This disease is hard to detect, while the most popular method for diagnosing is based on bone scintigraphy. This technology is based on scanning the whole body; however, the hot spots that are presented in the scanned image may mislead the results. Therefore, the accuracy of this methodology is not enough and makes the diagnosis of bone metastasis a real challenge. Thus, the researchers have been encouraged to increase the accuracy of diagnosing this disease by using machine learning and deep learning techniques. In this chapter, we present a comparative study for evaluating the performance of employing two deep learning techniques in this research direction. We selected the long-term recurrent convolutional network (LB-FCN, which stands for light-weighted bilinear fully convolutional network) and contrastive learning since they are not sufficiently evaluated in previous related works. The results have been reported by using six evaluation metrics: precision, recall, F1-score, sensitivity, specificity, and accuracy. The results show a demonstration of contrastive learning over LB-FCN.