Alqahtani Mohammed S, Abbas Mohamed, Alqahtani Ali, Alshahrani Mohammad, Alkulib Abdulhadi, Alelyani Magbool, Almarhaby Awad, Alsabaani Abdullah
Department of Radiological Sciences, College of Applied Medical Sciences, King Khalid University, Abha 61421, Saudi Arabia.
BioImaging Unit, Space Research Centre, Department of Physics and Astronomy, University of Leicester, Leicester LE1 7RH, UK.
Diagnostics (Basel). 2021 May 10;11(5):855. doi: 10.3390/diagnostics11050855.
Since late 2019, Coronavirus Disease 2019 (COVID-19) has spread all over the world. The disease is highly contagious, and it may lead to acute respiratory distress (ARD). Medical imaging can play an important role in classifying, detecting, and measuring the severity of the virus. This study aims to provide a novel auto-detection tool that can detect abnormal changes in conventional X-ray images for confirmed COVID-19 cases. X-ray images from patients diagnosed with COVID-19 were converted into 19 different colored layers. Each layer represented objects with similar contrast that could be defined as a specific color. The objects with similar contrasts were formed in a single layer. All the objects from all the layers were extracted as a single-color image. Based on the differentiation of colors, the prototype model was able to recognize a wide spectrum of abnormal changes in the image texture. This was true even if there was minimal variation of the contrast values of the detected uncleared abnormalities. The results indicate that the proposed novel method can detect and determine the degree of lung infection from COVID-19 with an accuracy of 91%, compared to the opinions of three experienced radiologists. The method can also efficiently determine the sites of infection and the severity of the disease by classifying the X-rays into five levels of severity. Thus, the proposed COVID-19 autodetection method can identify locations and indicate the degree of severity of the disease by comparing affected tissue with healthy tissue, and it can predict where the disease may spread.
自2019年末以来，新型冠状病毒肺炎（COVID-19）已在全球蔓延。该疾病具有高度传染性，可能导致急性呼吸窘迫（ARD）。医学成像在对该病毒进行分类、检测和评估严重程度方面可发挥重要作用。本研究旨在提供一种新型自动检测工具，能够检测确诊COVID-19病例的传统X射线图像中的异常变化。将确诊为COVID-19患者的X射线图像转换为19种不同颜色的图层。每个图层代表具有相似对比度的物体，可被定义为特定颜色。具有相似对比度的物体在单个图层中形成。所有图层中的所有物体被提取为单色图像。基于颜色差异，原型模型能够识别图像纹理中的广泛异常变化。即使检测到的未清除异常的对比度值变化很小，情况也是如此。结果表明，与三位经验丰富的放射科医生的判断相比，所提出的新方法能够以91%的准确率检测并确定COVID-19引起的肺部感染程度。该方法还可以通过将X射线分为五个严重程度级别，有效地确定感染部位和疾病严重程度。因此，所提出的COVID-19自动检测方法可以通过将受影响组织与健康组织进行比较来识别疾病位置并指示严重程度，并且可以预测疾病可能蔓延的部位。
