AI-Powered ADHD Diagnostic Support System

Currently, the prognosis of critically ill patients in the intensive care unit (ICU) is often assessed using the APACHE or SOFA scoring systems to predict mortality rates. However, these systems have been in use for over 20 years, and the data they are based on is outdated. Today, clinical data is more abundant and precise, and advancements in machine analysis capabilities have made traditional scoring systems insufficient for meeting the demands of accurate predictions and personalized medicine. To address this, our team has developed a survival prediction device for critically ill patients. Using an AI algorithm model, it utilizes clinical data such as physiological, laboratory, and examination results to predict the patient's survival rates for the next 30, 60, and 90 days, providing a more accurate reference for prognosis assessment and medical resource allocation.

The "Dosage Evaluation Method" applies AI technology to assess dosages for high-risk medications, using Vancomycin as an example. This method addresses limitations in clinical practice, where dosages are typically determined by physician experience or population pharmacokinetic parameters, which may not account for individual patient variability. Although several AI-based prediction models for dosage evaluation currently exist, most of these models directly suggest dosages, often without fully considering physician expertise. The innovation of this method lies in its approach to validating only the physician-selected dosage, providing reference feedback without influencing the physician’s final decision. This design not only respects the physician’s professional judgment but also enhances the safety of medication administration.

Our product combines acoustic technology with artificial intelligence to offer an innovative device focused on muscle quality assessment. Traditional muscle testing methods often rely on the subjective experience of physicians or expensive equipment such as CT scans and MRIs, which are not only costly but also cumbersome. Our device uses non-invasive acoustic signal analysis, combined with AI models, to provide fast and accurate diagnoses, significantly reducing medical costs and improving diagnostic efficiency. The product is lightweight and portable, allowing physicians to use it flexibly in clinics, communities, or home visits, making it suitable for diverse scenarios such as long-term chronic disease patients, athletes, and the elderly. With high-precision AI algorithms, the device can instantly provide analysis reports on muscle condition, such as strength, elasticity, and power, helping physicians make more scientifically-based clinical decisions.

This product addresses issues faced by physicians in assessing fistula blockage, including the lack of quantifiable measurements, high costs of precise assessments, and lengthy measurement times. By using patented audio equipment and a proprietary AI model, this product enables accurate assessment of fistula blockage levels, transforming what was once a time-consuming, labor-intensive, and experience-dependent task into an efficient and reliable process.

Our team has built predictive models based on whole transcriptome data from patients with advanced metastatic prostate cancer, focusing on two different mechanisms of action: androgen receptor antagonists and androgen synthesis inhibitors. These two types of drugs essentially offer only one chance of selection, as resistance to one renders the other ineffective. Using weighted gene correlation network analysis, we identified several gene modules correlated to specific drug responses. Then, we combined several machine-learning algorithms and constructed classification models for different drug mechanisms. Our goal is to use detection services to determine the most effective drug for patients before administration.