X光片肺部腫瘤電腦輔助診斷系統

Image Diagnosis System for Lung Nodule Detection

陳昶男(Chung-Nan Chen) ； 吳祖龍(Tsu-Lung Wu) ； 謝宗佑(Tsung-Yu Hsieh) ； 張士旭(Shih-Hsu Chang)

《中南盟研究年刊》 3卷1期 (2010/03) Pp. 45-54

https://doi.org/10.6850/JCSMA.201003_3(1).0005

摘要

目的：肺癌(lung cancer)的死亡率超過90%，而肺癌初期在X光片上只是一個可視的小圓點被稱為結節(nodule)。若能早期從X光影像上發現，早期治療，可提高患者的存活率。然而對於人體內部不易偵測的腫瘤及組織異常，就必需靠X光影像或腦斷層掃瞄(CT)來輔助增加診斷機率。但X光片偵測是否有小結節不是件容易的事，且其在成像時會有一些雜訊存在或是因眼睛疲勞都會導致誤判，影響治療時機。因此本研究利用數位影像處理技術及分析技術，將X光影像進行增強處理將可疑結節的影像增強出來，以協助醫師找出影像中之可疑區做進一步確認，預期提高診斷效率。方法：本研究對肺部X光片進行影像處理，進而以人工判圖方法，圈選可疑腫瘤區域。影像處理的方法有，1.利用差值影像做影像銳化；2.影像對比加強法（包括直方等化法、乘冪加強法）；3.平滑濾波法（包括高斯平滑濾波法、均值濾波法、平均濾波法）；4.邊緣偵測（包括一次階導、拉氏邊測、Sobel、Roberts等邊測法）；5.胸腔脊柱定位。結果：經本研究所提出的影像處理後，X光片的腫瘤會被凸顯。結論：實驗顯示，胸腔X光片經增強處理後，對於胸腔腫瘤的診斷的確有幫助。此外系統亦提供簡明的操作介面與快速的影像處理程式，來提高實際操作的實用性。

關鍵字

lung tumor ； lung nodule ； 影像處理 ； 肺癌 ； 肺腫瘤

並列摘要

Lung cancer is the primary importance in the death toll of malignant tumor announced on the webpage of Department of Health, Executive Yuan in Taiwan. The most common techniques for lung tumors detection used inc1ude chest radiography, and Computerized Tomography scans (CT). Although CT is more sensitive and precise technique, chest radiology (X-rays image) remains the initial and most common procedure since it is the most cost-effective. It is very important to improve doctor's diagnosis exactness in the lung disease X-rays image. The purpose of this project is to create a Computer Aided Diagnosis (CAD) System to process the X-rays images in order to extract and create information useful for radiologists during their decision making process.
The system implements several image processing techniques processing on X-rays images. These techniques inc1ude (1) histogram equalization, (2) image Gaussian smoothing, (3) median fi1ter, (4) first order derivative edge detection, and (5) uti1ize the difference image to shape the image. In addition the system also offers concise Graph User Interface (GUI) operations to help radiologists to process the X-rays images.
The database of Japanese Society of Radiological Technology (JSRT) is a standard database containing a total of 247 radiographs: 154 containing lung nodules of different diameter and 93 of patients with no disease. This database contains images with different characteristic of brightness; moreover some images may be affected by noise due to the imaging system. The experimental results show that the proposed system is useful for radiologists during their decision process.

並列關鍵字

medical images ； lung nodule ； lung Ca ； Computer Aided Diagnosis