Summary

    Chronic diseases have always been a vital healthcare issue worldwide. In this project, we are focused on developing and deploying a computer-aided diagnosis (CAD) system, especially powered by artificial intelligence (AI) and deep learning, for automatically grading, examining, and supporting to diagnose chronic diseases.

Advantages

    This project uses deep learning to develop the diagnosis, grading, and prediction systems of chronic diseases on the medical image of different organs or tissues and examine chronic diseases through different modality images to achieve personalized medicine for patients. The CAD system can provide recommendations applicable to patients for prognosis and medication of treatment guidelines. In addition, the topic will not only be based on the deep learning architecture but also on the characteristics of the target diseases to improve the performance of the CAD system.

Applications

(1) Diagnosis and Examination of Hyperthyroidism
    Hyperthyroidism is a chronic disease that causes cardiovascular disease, gastrointestinal upset, and nervous system problems and regularly requires long-term treatment and follow-up. Free tetraiodothyronine (fT4) may indicate an overactive thyroid is usually obtained and leveled from blood tests. Although fT4 can determine thyroid status, the blood test is time-consuming. Ultrasound is a commonly used modality for thyroid examination within the neck, taking multiple images to determine the structure and state of the thyroid. We expect to develop a computer-aided system to automatically select the best image interpreted by physicians from multiple neck ultrasound images and provide the level of fT4 information for follow-up and prognosis evaluation.

(2) Diagnosis and Examination of Chronic Obstructive Pulmonary Disease
    Chronic obstructive pulmonary disease (COPD) is an inflammatory lung disease that results from a prolonged time constant for lung emptying, caused by increased resistance of the small conducting airways and increased lung compliance as a result of emphysematous destruction. According to WHO, COPD has 6% of the global prevalence rate, about 16% among people over 40 years old in Taiwan. The current method to detect COPD is by examining the rate of expiratory airflow obstruction by pulmonary function, but that is not a good tool for assessing symptoms. We expect to create a deep learning system on chest X-ray images to diagnose COPD. This system can provide proper treatment and prognosis with post-diagnosis medication and related medical health information to achieve long-distance medical care for chronic disease.

Keywords

Chronic Disease, Artificial Intelligence, CAD System, Hyperthyroidism, COPD

◎ PI

PI Ruey-Feng Chang

Professor, Dept. of Computer Science and Information Engineering / BEBI, NTU

WebSite