CD4+ Lymphocyte counting using pixel-based features in low resolution fluorescence and bright field images

การนับลิมโฟไซต์ชนิด CD4+ โดยใช้ลักษณะเด่นของพิกเซลในภาพ ฟลูออเรสเซนส์และภาพพื้นหลังสว่างความละเอียดต่ำ

Name: Usanee Apijuntarangoon

ThaSH: Lymphocyte

Classification :.DDC: 616.079

ThaSH: Fluorescence

ThaSH: Leucocytes

Abstract: Recently, automatic image-based CD4+ lymphocyte counting instrument is available in the market. This instrumentation consists of a portable bench-top, which cheaper than the standard counting machine, flow cytometer. CD4+ lymphocytes are counted via the bi-color fluorescence image. However, counting cell using only fluorescence image has certain limitations since no guarantee that the color spots are cells. In this study, this problem is solved by using a corresponding bright field image to guarantee the reliability of detection method. CD4+ lymphocytes are stained by immunofluorescence technique and they can be seen in both green and red fluorescence image. The overlapped region between the white blood cells (WBCs) in bright field image and color spots on corresponding green and red fluorescence images are CD4+ lymphocyte. We proposed the automatic unsupervised method for cell detection in bright field and fluorescence images in order to count CD4+ lymphocytes. The WBCs detection method in bright field image was developed based on the cells morphology which represents the outstanding edge and intensity information. The segmentation was done twice in order to reduce false detection caused by background. Besides, the classical method for separating adhere cells was implement. Then, the roundness was used to select the WBCs based on shape. The WBCs detection approach gave a good sensitivity and positive predictive value (PPV). For cells detection in fluorescence approach, we used a multi-scale top-hat transform as an intensity enhancement. The thresholding values from 0 to 255 were applied to generate ROC curve in order to evaluate the performance of the CD4+ lymphocyte detection algorithm. The obtained ROC curves of green and red fluorescence images show the high true positive rate (TPR) while low false positive rate (FPR). The best thresholding value from each ROC was applied in order to evaluate the counting performance and we found out that counting result shows a good sensitivity and PPV. However, the accuracy should be further improved for the real medical application.

Chiang Mai University Library

Address: CHIANG MAI

Email: cmulibref@cmu.ac.th

Name: Yuttana Mundee

Role: Chairman

Name: Nipon Theera-Umpon

Role: Member

Name: Chatchai Tayapiwatana

Role: Member

Name: Sansanee Auephanwiriyakul

Role: Member

Created: 2015

Modified: 2017-09-17

Issued: 2017-09-17

วิทยานิพนธ์/Thesis

application/pdf

CallNumber: Th 616.079 U84C

eng

DegreeName: Master of Engineering

Level: Master's Degree

Descipline: Biomedical Engineering

Grantor: Chiang Mai University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	enbio60715uap_full.pdf	2.58 MB

ใช้เวลา

0.04103 วินาที