A computer-aided diagnostic system for personal health using vision-based sensors

In the most recent years, the health care system in Hong Kong has serious problems with the workforce shortage and elderly population growth. To maintain a high throughput of the diagnostic performance in the present medical system, the doctors are having much longer working hours than before. In such a situation, high pressure and high workload working environment might reduce the precision of marking medical decisions.To solve the above issues, we developed a computer-aided diagnostic system by using the newest vision-based sensors and artificial intelligence technologies. The system developed could largely reduce the doctors' workload and highly improve diagnostic efficiency.In the application of the vision-based sensors, we developed an easy-touse multiple sensors calibration system. Such a system has the ability to figure out the geometric relationships between the sensors using the point cloud data. All the data can be integrated together with the same coordination system representation. Meanwhile, we further developed a conversion method to convert the 3D point cloud data to 2D images. By applying this conversion method, the well-established 2D image processing methods can be used to process the point cloud data.In the mental disease recognition system, we built up a Hong Kong mental disease feature database. The facial expression data from the patients were captured by the sensors during the medical diagnostic interviews. We additionally developed a recognition system using deep neural networks to process and recognize mental diseases based on the machine learning approach. The system developed in this project can distinguish the omega sign from a control case. Omega sign is an important sign shown on the human face that indicates depression. The final results illustrated that our system can successfully achieve a significant high accuracy with a limited size of the supplied dataset.香港的醫療系統人手短缺的問題日趨嚴重，加上香港的老年人口比率日益擴大，本港醫療系統正在承受極大的壓力。為了維持醫療系統在診症方便的高吞吐量，本港醫生的工作時數一向普遍較正常為長。長時間高壓和高工作量的環境下工作，醫生的判斷能力被受考驗。長此下去，本港的醫療系統難以維持高質素的服務水準。為了減輕醫生的工作量和提升診症的效率，我們希望運用最新的視覺感應器和人工智能技術，開發出一套電腦輔助的自診斷系統。在可以輔助醫生診症和判斷病人病情的同時，希望可以提升大眾對個人健康的關注和整體的健康水平。在本論文內，我們實作出一套運用視覺感應器來診斷心理疾病的系統和輔助醫生診症。在感應器應用上，我們開發出一套簡易使用的多感應器校準系統。此系統利各感應器所取得點雲（Point Cloud） 數據，來找出不同感應器之間的相對位置，並把這些感應器的數據融合在同一個坐標系統上。從而我們可以得出一個更全面的點雲數據。與此同時，我們提出了一個把三維的點雲數據轉換為二維相片的方法。希望通過此轉換方式，把現階段成熟的相片辨識技術運用在處理點雲數據上。在心理疾病的辨識上， 我們創立了香港心理疾病的特徵數據庫。在醫生診症的同時， 運用感應器來收集病人的面容數據。與此同時，我們開發出一個採用了機械學習技術的心理疾病辨識系統。此辨識系統通過辨認病人面上的奧米加標記（omega sign） 來找出病人是否有如抑鬱症等的心理疾病。此系統在數據稀少的情況下， 也達到了比較高的辨識準確率。Lo, Sheung Lai.M.Phil. Chinese University of Hong Kong 2021.Includes bibliographical references (leaves ).Abstracts also in Chinese.Title from PDF title page (viewed on …)