Healthcare is one of the most important services to preserve the quality of our daily lives, and it is capable of dealing with issues such as global aging, increase in the healthcare cost, and changes to the medical paradigm, i.e., from the in-facility cure to the prevention and cure outside the facility. Accordingly, there has been growing interest in the smart and personalized healthcare systems to diagnose and care themselves. Such systems are capable of providing facility-level diagnosis services by using smart devices (e.g., smartphones, smart watches, and smart glasses). However, in realizing the smart healthcare systems, it is very difficult, albeit impossible, to directly integrate high-precision healthcare technologies or scientific theories into the smart devices due to the stringent limitations in the computing power and battery lifetime, as well as environmental constraints. In this dissertation, we propose three optimization methods in the field of cell counting systems and gait-aid systems for Parkinson's disease patients that address the problems that arise when integrating a specialized healthcare system used in the facilities into mobile or wearable devices. First, we present an optimized cell counting algorithm based on heuristic optimization, which is a key building block for realizing the mobile point-of-care platforms. Second, we develop a learning-based cell counting algorithm that guarantees high performance and efficiency despite the existence of blurry cells due to out-focus and varying brightness of background caused by the limitation of lenses free in-line holographic apparatus. Finally, we propose smart gait-aid glasses for Parkinson’s disease patients based on mathematical optimization. ⓒ 2017 DGIST
Table Of Contents
I. Introduction 1--
1.1 Global Healthcare Trends 1--
1.2 Smart Healthcare System 2--
1.3 Benefits of Smart Healthcare System 3--
1.4 Challenges of Smart Healthcare. 4--
1.5 Optimization 6--
1.6 Aims of the Dissertation 7--
1.7 Dissertation Organization 8--
II.Optimization of a cell counting algorithm for mobile point-of-care testing platforms 9--
2.1 Introduction 9--
2.2 Materials and Methods. 13--
2.2.1 Experimental Setup. 13--
2.2.2 Overview of Cell Counting. 16--
2.2.3 Cell Library Optimization. 18--
2.2.4 NCC Approximation. 20--
2.3 Results 21--
2.3.1 Cell Library Optimization. 21--
2.3.2 NCC Approximation. 23--
2.3.3 Measurement Using an Android Device. 28--
2.4 Summary 32--
III.Human-level Blood Cell Counting System using NCC-Deep learning algorithm on Lens-free Shadow Image. 33--
3.1 Introduction 33--
3.2 Cell Counting Architecture 36--
3.3 Methods 37--
3.3.1 Candidate Point Selection based on NCC. 37--
3.3.2 Reliable Cell Counting using CNN. 40--
3.4 Results 43--
3.4.1 Subjects . 43--
3.4.2 Evaluation for the cropped cell image 44--
3.4.3 Evaluation on the blood sample image 46--
3.4.4 Elapsed-time evaluation 50--
3.5 Summary 50--
IV.Smart Gait-Aid Glasses for Parkinson’s Disease Patients 52--