Biomedical image processing by using image and signal processing

การประมวลผลภาพทางชีวการแพทย์โดยใช้ภาพและการประมวลผลสัญญาณ

Name: Pikul Vejjanugraha

keyword: Fundus image analysis

; Cup-to-disc ratio

; Support vector machine

; 3D active contour model

; Inhomogeneous motion analysis

; Velocity vector maps

Abstract: Biomedical image processing is to integrate the medical knowledge to the digital imaging techniques for medical diagnostics. The prior knowledge of the specific medical content of images is needed to realize. Generally, it covers four main areas: Image formation, Image Visualization, Image Analysis, and Image Management. In this research, Image enhancement and Image Analysis are mainly focused. Two case studies with different types of image formation and prior knowledge are investigated. Case study 1 is an analysis of a retinal image of human eyes. We introduce an automatic screening technique for glaucoma diagnosis using Support Vector Machine (SVM). The optic nerve head (ONH) is considered as a target object for segmentation, consists of the optic disc (OD) and cup (OC). The important feature called Cup-to-Disc ratio (CDR) is calculated from the diameters of the OD and OC and it is considered at 0.65 as a threshold for the mass screening process. For image preprocessing and enhancement, the region of interest (ROI), low-level, and high-level image processing are processed in order to enhance the color information of OD and OC on RGB bands, and the pixel-based clustering such as k-mean clustering is applied to reduce unwanted noises. The unsmoothed edges of OD and OC are solved by ellipse fitting. The shape information of OD and OC is then collected to classify glaucoma’s stages. For the analyzing process, a computational intelligence-based classifier such as an artificial Neural Network (aNN) and a Support Vector Machine (SVM) is used to classify the stages of glaucoma as follows normal, suspect, and abnormal classes. The unbalanced decision tree (UDT) and one-vs-the-rest (OVR) techniques are combined to overcome the limitation of SVM. Finally, it is found that the classification model with both UDT and OVR techniques yielded the most reliable result. Case Studies 2 to 4 are analysis and visualization of 2D and 4D chest CT-scan images for pulmonary disease diagnosis. It is a 3D model-based analysis by using the Active Contour Model (ACM). This case study introduces the new approach to evaluate the velocity vectors of the lung motion by using 3D ACM and learn the inhomogeneous motion pattern of each lung lobe to generate the predictive model. The non-rigid registration model using its biophysical model is applied. The velocity vectors between End Inspiratory (EI) and End Expiratory (EE) models are evaluated by the corresponding points on the parametric surface model of the EI and EE models. The external energy from the EI models is the external force that pushes the 3D parametric surface reaching the boundary. The external forces such as balloon and Gradient Vector Flow (GVF) were adjusted adaptively based on the Zratio which calculated from the ratio of the maximum value of EI to the EE model in the Z-axis. Next, the feature representation is studied and evaluated based on the lung structure separated into five lobes. To screening the lung diseases into the normal, obstructive lung, and restrictive lung, stepwise regression, and Artificial Neural Network technique are used to evaluate the result. In conclusion, the inhomogeneous motion pattern of lungs integrated with the medical-based knowledge can be used to analyze the lung diseases: firstly, by differentiating normal and inhomogeneous motion pattern, secondly by separating restrictive and obstructive lung diseases, and thirdly basing on the cause and location of the disease which is the function of the immune and lymphatic system

Thammasat University. Thammasat University Library

Address: BANGKOK

Email: preserv@tu.ac.th

Name: Waree Kongprawechnon

Role: advisor

Name: Kotani, Kazunori

Role: co-advisor

Created: 2019

Modified: 2023-01-26

Issued: 2023-01-26

วิทยานิพนธ์/Thesis

application/pdf

DOI: 10.14457/TU.the.2019.1644

eng

DegreeName: Doctor of Philosophy

Level: Doctoral Degree

Descipline: Engineering and Technology

Grantor: Thammasat University

©copyrights Thammasat University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	10186pikul.pdf	3.43 MB

ใช้เวลา

0.029353 วินาที