基於深度學習之高通量顯微影像系統於血液分析;Deep learning based high-throughput microscope imaging system for blood cell analysis

NCU Institutional Repository > 生醫理工學院 > 生物醫學工程研究所 > 博碩士論文 > Item 987654321/85785

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/85785

題名:	基於深度學習之高通量顯微影像系統於血液分析;Deep learning based high-throughput microscope imaging system for blood cell analysis
作者:	楊璨華;Yang, Tsan-Hua
貢獻者:	生物醫學工程研究所
關鍵詞:	中性粒細胞;菲涅耳;全血分析;骨髓抑制
日期:	2021-09-08
上傳時間:	2021-12-07 11:24:39 (UTC+8)
出版者:	國立中央大學
摘要:	治療癌症病患最常見的方法之一為化學治療，但常有發生骨髓抑制狀況，使患者血球細胞數量減少，最常見的原因是化學藥物減緩幹細胞和特化後代的分裂繁殖能力，導致造血功能無法正常運作，當嗜中性白血球數量減少會使癌症病患受到全身性感染的機率增加，目前的治療手段有預防性抗菌藥物和造血生長因子補充劑，同時降低化療劑量，因此限制了化學治療的功效甚至導致化療失敗，反覆感染更是會導致住院時間延長、治療延遲的問題。本研究中使用菲涅耳數位全像術(Fresnel Digital Holography, FDH)原理研製一套血液分析系統，利用純量繞射理論簡化了光學成像設備，不再需要龐大且複雜的光學元件，沒有了透鏡限制視場(Field of view, FOV)大小，可同時具備廣視場(30mm2)並達到接近繞射極限之高空間解析度，透過控制光源的空間相干性在傳感器上記錄繞射影像，經由深度神經網路(Deep Neural Network, DNN)分類計數血球繞射圖像，並使用微型電腦做為系統的控制中樞，建立不受場域限制且可即時進行血液分析之高通量顯微系統。在本篇論文中以全血細胞分析監測血球數目，再搭配自主設計微流道採血晶片和深度學習網路後，目前對於三類常見血球細胞，如:紅血球、白細胞及血小板之計數準確率分別可達到92%、75%與78%。實驗結果驗證本系統只需微量血液即可進行快速、大面積的血液細胞表徵分析與全血計數功能，可即時反映因血球數異常而引起的各種適應症，可望在未來進一步細分並計數各類白血球，達到即時檢測有無中性粒細胞減少症，大幅度降低病患全身性感染的風險，從而減少住院需求並避免療程的延遲。 ;The occurrence of bone marrow suppression will reduce the number of blood cells. The Mmost common reason is that chemical drugs slow down the division and reproduction of stem cells and specialized offspring, causing the hematopoietic function to fail to function properly., Wwhen the number of neutrophils decreases, cancer patients are more likely to suffer systemic infections. The current treatment methods include preventive antibacterial drugs and hematopoietic growth factor supplements, and . Tthe need to reduce the dose of chemotherapy, thus limiting the treatment efficacyefficacy of treatment, infecting repeatedly will lead to prolonged hospital stays and delayed treatment. In my research, a blood analyzer was established using the principle of Fresnel digital holography(FDH). We simplify the optical imaging equipment by scalar diffraction theory, and no large and complex optical components are needed. Without the lens to limit the field of view(FOV) size, it can simultaneously have a wide field of view (30mm^2 ) and reach a high spatial resolution close to the diffraction limit. We recorded the diffraction image on the sensor by controlling the spatial coherence of the light source, and the diffraction image of the blood cell is classified and counted by Deep Neural Network(DNN), and the microcomputer is used as the control center of the system. The system is a high-throughput microscopy system without field-limited, and it can analyze blood in real-time. In this paper, we use the whole blood cell analysis to monitor the number of blood cells, ; coupled with the self-designed microfluidic chip and deep learning network, the accuracy of red blood and white blood cell counts can reach 9281% to 7583% . The experimental verification results confirm that the system can perform rapid and large-area blood cell characterization analysis and complete blood count function with only a small amount of blood, which can respond to various indications caused by abnormal blood counts. In the future, we will further count various types of white blood cells to achieve real-time detection of neutropenia, which greatly reduces the risk of systemic infection in patients, thereby reducing the needs ofneed for hospitalization and avoiding delays in treatment.
顯示於類別:	[生物醫學工程研究所 ] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	82	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....