[Taichung News] Kidney cancer is one of the most common malignant tumors globally, and distinguishing between benign and malignant renal tumors has long been a challenge for the medical community. To improve diagnostic accuracy, Dr. Wei-Shuo Wu, recipient of the Outstanding Physician Award from China Medical University Hospital, collaborated with Section Chief Chih-Jen Tsai and his research team from the Precision Health Research Center at Asia University. Utilizing deep learning technology, they developed a model to automatically classify the benignity or malignancy of tumors in renal ultrasound images. This research was recognized by the technical committee at the 2024 International Conference on Computational Intelligence, Biomedicine and Medical Imaging (ICIIBMS 2024), where Dr. Wu delivered an online oral presentation to showcase the latest findings.  

 Deep Learning Empowers Medical Image Interpretation to Enhance Diagnostic Accuracy  

The research team conducted deep learning training and testing on 880 renal ultrasound images, using Convolutional Neural Networks (CNN) to automatically differentiate between benign and malignant tumors. During the experiments, the team compared ten different CNN models, including VGG16, ResNet18, and DenseNet121, and employed transfer learning techniques to optimize model performance. The results indicated that the VGG16 model achieved the best diagnostic performance, with a sensitivity of 79% and a specificity of 86%, demonstrating the potential of artificial intelligence in the field of medical imaging diagnostics.  

 Research Breakthrough: Supporting Clinical Decisions and Reducing Misdiagnosis Risks  

Currently, ultrasound image diagnosis relies heavily on the experience and skill of the physician; therefore, in regions with limited medical resources, diagnostic accuracy may be compromised due to a lack of experienced technicians. Through this study, Dr. Wei-Shuo Wu's team successfully developed an automated classification system based on deep learning. This system not only assists physicians in early diagnosis but also reduces misdiagnosis rates, avoiding unnecessary invasive examinations or surgeries.  

Furthermore, the study found that the five-year survival rate for early-stage kidney cancer is as high as 92.6%, whereas the survival rate for late-stage patients is only 11.7%. Consequently, using AI technology to detect the benignity or malignancy of tumors early has a significant impact on patient treatment decisions and prognosis.  

 International Recognition: Research Findings Presented at EI Conference in Japan  

The results of this research were presented at the ICIIBMS 2024 international conference. This EI-indexed conference covers cutting-edge technologies such as biomedicine, artificial intelligence, and medical imaging, attracting experts and scholars from around the world. Dr. Wei-Shuo Wu personally delivered an online oral presentation at the conference, detailing the research methodology and results, which received high praise from international experts. This research not only demonstrates Taiwan's innovative strength in the field of AI healthcare but also provides a new solution for the early diagnosis of kidney cancer.  

 Future Outlook: AI-Assisted Diagnosis as a New Trend in Healthcare  

With the continuous advancement of artificial intelligence, the application of deep learning in medical image analysis will become increasingly widespread. Chih-Jen Tsai's team plans to further expand the scale of their research, aiming to improve model accuracy through training on larger datasets and promoting the clinical application of AI-assisted diagnostic systems, ensuring that intelligent medical technology can be truly implemented to benefit more patients.