卷機類神經網路實作：機器學習肋骨骨折的胸部X光線平片影像辨識

Taipei Medical University Institutional Repository > 醫學院 > 人工智慧醫療碩士在職專班 > 博碩士論文 > Item 987654321/61659

請使用永久網址來引用或連結此文件: http://libir.tmu.edu.tw/handle/987654321/61659

題名:	卷機類神經網路實作：機器學習肋骨骨折的胸部X光線平片影像辨識 Convolutional Neural Network for Recognition in Rib Fracture from Radiographic Plain Film
作者:	江紀明 CHIANG, CHI-MING
貢獻者:	醫學院人工智慧醫療碩士在職專班康峻宏
關鍵詞:	肋骨骨折;深度學習;卷機類神經網路 rib fracture;deep learning;Convolutional Neural Network
日期:	2021-07-14
上傳時間:	2022-04-28 23:01:49 (UTC+8)
摘要:	肋骨骨折在胸部創傷中約佔61%至90%，X光線平片是檢查的首選方法。但是，往往因疼痛後的非精確的檢查擺位，造成投照體位與角度，甚或是X光線曝光條件選擇的不當，而造成臨床醫師判斷的困難而造成漏診。藉由胸部電腦斷層的輔助，可提高約4倍的肋骨骨折確診率；X光線平片的漏診率為5.91%，胸部電腦斷層的漏診率為1.63%。然而，三維電腦斷層的醫療成本較高，若是能以人工智慧對X光線平片進行影像辨識，初步篩選出肋骨骨折疑似病例後，再輔以胸部電腦斷層確診，既可降低醫療成本；亦可減低醫師負擔。因此本論文欲藉由深度學習之卷機類神經網路的研究方法，對胸部X光線平片進行肋骨骨折的影像辨識。就健康經濟學的層面而言，醫師作決策的方法是找出最小成本的方法，並在此選項下保證一定的準確率。而所謂的成本同時包括醫療檢查花費及後續治療成本、以及肋骨骨折的漏診與病患失能所造成的等值貨幣損失、醫師因漏診而承擔的醫療訴訟賠償、與病患失去未受傷前之工作收入等。綜上所述，吾等預期的成果為:藉由本論文所提供之研究方法，降低肋骨骨折之醫療診斷成本與醫師影像判讀的負擔。 Fractures of the ribs account for about 61% to 90% of chest trauma, and plain film in radiography are the preferred method of examination in the first place. However, the inaccurate positioning in the examination caused by the pain often results in improper selection of the position and angle of the projection, or even the improper selection of the X-ray exposure conditions. With the assistance of chest computer tomography, the diagnosis rate of rib fractures can be increased by about 4 times; the missed diagnosis rate of plain X-ray films is 5.91%, and the missed diagnosis rate of chest computer tomography is 1.63%. However, the medical cost of 3D computer tomography is relatively high. If it is possible to use artificial intelligence to identify X-ray plain films in suspected cases of rib fractures for initially screen, then supplemented with chest computer tomography to confirm the diagnosis can reduce medical costs. Moreover, it also can reduce the burden on physicians. Therefore, this thesis intends to use the research method of the deep learning (Convolutional Neural Network) to identify the rib fracture image on the chest X-ray plain film. In terms of health economics, the way doctors make decisions is to find the option with the lowest cost, and to ensure a certain threshold of diagnosis under this option. The so-called cost includes the cost of resources for medical examinations for diagnosis and follow-up treatment, as well as the equivalent of monetary loss caused by missed diagnosis of rib fractures and patient disability; medical lawsuit compensation and patient loss caused by missed diagnosis. In summary, the expected result of this theory is to reduce the cost of medical diagnosis of rib fractures and the burden of physicians' interpretation of images by the research methods provided by this theory.
描述:	碩士指導教授：康峻宏委員：彭徐鈞委員：陳右潁
資料類型:	thesis
顯示於類別:	[人工智慧醫療碩士在職專班] 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	164	檢視/開啟

檢視Licence

在TMUIR中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....