電腦輔助偵測系統於阿茲海默症應用分析

The Development of Computer-Aided Detection System for Alzheimer's Disease

蔡育臻(Yu-Chen Tsai)

指導教授 ： 蘇振隆

中原大學/工學院/生物醫學工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000547

摘要

台灣失智症盛行率逐年上升，預估2018年到2031年失智症人口將由27萬上升至46萬人，將帶來嚴重衝擊。失智症最常見的為阿茲海默症，目前醫師診斷仍多以臨床智能量表(CDR)及問診方式較為多，但藉由兩項診斷方式較為主觀，也可能延誤治療。若能利用已有之CT影像來檢視腦部是否有萎縮的現象，可縮短病人排檢及檢查時間，達成及早發現與治療。
本研究主要是探討先前研究室所開發的電腦輔助偵測系統於不同醫院之阿茲海默症病人探討其適用性。以訓練新閾值並進行系統效能評估，提供不同醫院的醫師未來判斷阿茲海默症嚴重程度之參考。研究步驟包含:(1)先以20組CT影像依原先設定之閾值進行腦組織容量比例的初步測試；(2)應用Image J來計算不同醫院腦實質組織與腦脊髓液的區塊平均灰階值，並以平均灰階值之比例來尋找電腦輔助系統的適用閾值；(3)訓練出20組CT影像之最佳閾值； (4)使用最佳閾值並利用30組測試組與病人之CDR影像進行系統效能評估。
結果顯示: (1)經由20組CT影像進行腦組織容量初步測試結果靈敏度50%、特異性0%、準確度25%，結果顯示使用原有系統閾值90，無法有效分割腦實質組織與腦脊髓液；(2)以訓練組20組訓練及公式推算最佳閾值73，使用30組測試組，並加入11項腦萎縮參數進行系統效能評估，其準確度為86.7%、靈敏度為86.2%、特異性為95.5%、Kappa值為0.821。若將測試組中排除經神經內科醫師確診為多發性腦梗塞之病例，再統計其效能，則準確度、靈敏度、特異性及Kappa值提升為89.7%、89.8%、96.4%及0.862。
本研究提出閾值調整之方法，此方法可產生因影像來源而改變系統之閾值，使系統可適用於不同之醫院環境。這不僅能讓醫師再判斷阿茲海默症方面能有更準確的評估，同時也能夠在評估腦萎縮程度方面能有更精確的分析並據以及早對患者進行治療。

關鍵字

阿茲海默症 ； 影像處理 ； 可調式之閾值 ； 電腦輔助偵測系統 ； 腦萎縮

並列摘要

The prevalence of dementia in Taiwan is increasing year by year, and the number of people with dementia expected to rise from 270,000 to 460,000 between 2018 and 2031, with a serious impact. The most common form of dementia is Alzheimer’s disease, and doctors are diagnosed based on Clinical Dementia Rating (CDR) and medical consultations, but the two diagnoses are more subjective and may delay treatment. If the existing CT images can used to examine the presence of atrophy in the brain, it can shorten the time of patient scans and examinations, and achieve early detection and treatment.

This study focuses on the applicability of computer-assisted detection systems previous developed by our laboratory in Alzheimer's patients in different hospitals. To train new thresholds and conduct systematic performance assessments, physicians in different hospitals can used to determine the severity of Alzheimer's disease in the future. The results will enable doctors in different hospitals to judge the severity of Alzheimer's disease. Research steps are as follows: (1) use 20 sets of CT images and the initial threshold to test the volume ratio of brain tissue. (2) Apply Image J to calculate the blocks of average grayscale value of brain parenchyma and cerebrospinal fluid in different hospitals, and use the ratio of the average grayscale value to find the suitable thresholds in computer-aided system. (3) Determine the optimal threshold of 20 sets of images. (4) Use the optimal threshold, 30 test groups images and corresponding CDR of patients to assess the performance evaluation system.

The results showed that: (1) the original system threshold of 90 could not effectively divide the brain's actual tissue and cerebrospinal fluid of brain tissue, and the sensitivity, specificity, and accuracy of this system are 50%, 0%, and 25%, respectively, for the initial test results of 20 sets of CT images. (2) After calculated, optimal threshold 73 is used to conduct the system performance evaluation, and the results showed that the accuracy, sensitivity, specificity, and Kappa values are 86.7%, 86.2%, 95.5%, and 0.821, respectively. If the test group excluded cases of multiple cerebral infarction diagnosed by neurologists and their effectiveness was counted, the accuracy, sensitivity, specificity and Kappa values were increased to 89.7%, 89.8%, 96.4% and 0.862, respectively.

In this study, a method of threshold adjustment proposed, which can produce a system threshold to change the system because of the image source, so that the system can be applied to different hospital environments. Not only will this allow physicians to evaluate Alzheimer's disease more accurately, but it will also allow for a more accurate analysis of the extent of brain atrophy and early treatment of patients.

並列關鍵字

CT images ； Adjustable threshold ； Computer-Aided Detection System ； Brain Atrophy

參考文獻

[1] 台灣失智症協會. 認識失智症. Available: http://www.tada2002.org.tw/tada_know_02.html, 2019/07/01資料擷取

[2] M. Castleman, 阿茲海默診療室: 天下文化,台北, 2001.

[3] KL Double, GM Halliday, JJ Krill, JA Harasty, K Cullen, WS Brooks, et al., "Topography of brain atrophy during normal aging and alzheimer's disease," Neurobiology of Aging, 17: 513-21, 1996.

[4] S Kim, YC Youn, GYR. Hsiung, SY Ha, KY Park, HW Shin, et al., "Voxel-based morphometric study of brain volume changes in patients with Alzheimer’s disease assessed according to the Clinical Dementia Rating score," J Clin Neurosci, 18(7):916-21, 2011.

[5] C Möller, H Vrenken, L Jiskoot, A Versteeg, F Barkhof, P Scheltens, et al., "Different patterns of gray matter atrophy in early- and late-onset Alzheimer’s disease," Neurobiology of Aging, 34:2014-22, 2013.