Deep learning has been increasingly applied to diagnostic dermatology. However, the absence of standards in dermatologic images as well as diverse artifacts make it difficult to achieve reliable performance, compatibility, and interpretability, and ultimately to have a practical use by clinicians. This project aims to develop and improve a deep learning algorithm for processing and diagnosis of dermatologic images. The state-of-the-arts technologies in deep learning will be exploited to resolve such problems.
Pet owners do not always have enough time to take care of their pets, but they need intelligent healthcare solutions to take place. We have developed an intelligent pet healthcare solution where pets' emotion and body language are jointly analyzed from videos using artificial intelligence (AI). The solution would lay a connerstone for innovation in animal healthcare industries.
The functional development of the fetal brain has not been scientifically elucidated. Our challenge was to answer this question by using in utero noninvasive magnetic resonance imaging (MRI). Using artificial intelligence (AI) technologies, we had developed not only a set of robust tools for MRI data processing and analyses dedicated to the moving fetus, but also diagnostic methods for fetal brain and placental impairments especially for congenital heart disease and fetal growth restriction.
How does the brain work when you sleep? A key to answer this question is to figure out that the resting state brain functions are 'fractal'. We had developed a novel fractal-based statistical methodology to quantify the resting state functional network of the brain through functional MRI. The fractal-based analytic methods have been successfully applied to study depression, autism, and cognitive functions in resting state.
Image and video content analyses by machine vision become more important in the era of the fourth industrial revolution. Based on machine learning, we have developed robust techniques for real-time multi-object detection and tracking in compressed videos whose computation could be dramatically accelerated without decoding while its performance was acceptably kept. The proposed methods were successfully applied to intelligent surveillance systems and interactive broadcasting services.
AI-based methodologies for diagnosing virus infection