Alqahtani Mohammed S, Abbas Mohamed A, Alqahtani Ali M, Alshahrani Mohammad Y, Alkulib Abdulhadi J, Alelyani Magbool A, Almarhaby Awad M
Department of Radiological Sciences, College of Applied Medical Sciences, King Khalid University, Abha, Saudi Arabia.
BioImaging Unit, Space Research Centre, Department of Physics and Astronomy, University of Leicester, Leicester, UK.
Curr Med Imaging. 2022;18(5):563-569. doi: 10.2174/1573405617666210910150119.
Since late 2019, Coronavirus Disease 2019 (COVID-19) has spread around the world. It has been determined that the disease is very contagious and can cause Acute Respiratory Distress (ARD). Medical imaging has the potential to help identify, detect, and quantify the severity of this infection. This work seeks to develop a novel auto-detection technique for verified COVID-19 cases that can detect aberrant alterations in traditional X-ray pictures.
Nineteen separately colored layers were created from X-ray scans of patients diagnosed with COVID-19. Each layer represents objects that have a similar contrast and can be represented by a single color. In a single layer, objects with similar contrasts are formed. A single color image was created by extracting all the objects from all the layers. The prototype model could recognize a wide range of abnormal changes in the image texture based on color differentiation. This was true even when the contrast values of the detected unclear abnormalities varied slightly.
The results indicate that the proposed novel method is 91% accurate in detecting and grading COVID-19 lung infections compared to the opinions of three experienced radiologists evaluating chest X-ray images. Additionally, the method can be used to determine the infection site and severity of the disease by categorizing X-rays into five severity levels.
By comparing affected tissue to healthy tissue, the proposed COVID-19 auto-detection method can identify locations and indicate the severity of the disease, as well as predict where the disease may spread.
自2019年末以来，2019冠状病毒病（COVID-19）已在全球传播。已确定该疾病具有很强的传染性，可导致急性呼吸窘迫（ARD）。医学成像有潜力帮助识别、检测和量化这种感染的严重程度。这项工作旨在开发一种针对确诊COVID-19病例的新型自动检测技术，该技术可检测传统X光片中的异常改变。
从确诊为COVID-19的患者的X光扫描中创建了19个颜色各异的图层。每个图层代表具有相似对比度且可用单一颜色表示的物体。在单个图层中，形成具有相似对比度的物体。通过从所有图层中提取所有物体创建了一幅单彩色图像。原型模型能够基于颜色差异识别图像纹理中的多种异常变化。即使检测到的不清晰异常的对比度值略有变化，情况也是如此。
结果表明，与三位经验丰富的放射科医生评估胸部X光图像的意见相比，所提出的新方法在检测和分级COVID-19肺部感染方面的准确率为91%。此外，该方法可通过将X光片分为五个严重程度级别来确定疾病的感染部位和严重程度。
通过将受影响组织与健康组织进行比较，所提出的COVID-19自动检测方法可以识别疾病位置、指示疾病严重程度，并预测疾病可能传播的部位。
