UreteroPelvic Junction Obstruction (UPJO) is a common hydronephrosis disease in children that can result in an even progressive loss of renal function. Ultrasonography is an economical, radiationless, noninvasive, and high noise preliminary diagnostic step for UPJO. Artificial intelligence has been widely applied to medical fields and can greatly assist doctors' diagnostic abilities. The demand for a highly secure network environment in transferring electronic medical data online, therefore, has led to the development of blockchain technology. In this study, we built and tested a framework that integrates a deep learning diagnosis model with blockchain technology. Our diagnosis model is a combination of an attention-based pyramid semantic segmentation network and a discrete wavelet transformation-processed residual classification network. We also compared the performance between benchmark models and our models. Our diagnosis model outperformed benchmarks on the segmentation task and classification task with$\mathsf{MloU}=87.93,\ \mathsf{MPA}=93.52$, and$\text{accuracy}=91.77\%$. For the blockchain system, we applied the InterPlanetary File System protocol to build a secure and private sharing environment. This framework can automatically grade the severity of UPJO using ultrasound images, guarantee secure medical data sharing, assist in doctors' diagnostic ability, relieve patients' burden, and provide technical support for future federated learning and linkage of the Internet of Medical Things (IoMT).
Muhammad Umar Nasir, Muhammad Zubair, Taher M. Ghazal, Muhammad Farhan Khan · 9 authors
Kidney cancer is a very dangerous and lethal cancerous disease caused by kidney tumors or by genetic renal disease, and very few patients survive because there is no method for early prediction of kidney cancer. Early prediction of kidney cancer helps doctors start proper therapy and treatment for the patients, preventing kidney tumors and renal transplantation. With the adaptation of artificial intelligence, automated tools empowered with different deep learning and machine learning algorithms can predict cancers. In this study, the proposed model used the Internet of Medical Things (IoMT)-based transfer learning technique with different deep learning algorithms to predict kidney cancer in its early stages, and for the patient's data security, the proposed model incorporates blockchain technology-based private clouds and transfer-learning trained models. To predict kidney cancer, the proposed model used biopsies of cancerous kidneys consisting of three classes. The proposed model achieved the highest training accuracy and prediction accuracy of 99.8% and 99.20%, respectively, empowered with data augmentation and without augmentation, and the proposed model achieved 93.75% prediction accuracy during validation. Transfer learning provides a promising framework with the combination of IoMT technologies and blockchain technology layers to enhance the diagnosing capabilities of kidney cancer.