Kanumuri Chalapathiraju, Chodavarapu Renu Madhavi
Electronics and Communication Engineering, S.R.K.R Engineering College, Bhimavaram, Andhra Pradesh India.
Electronics and Instrumentation Engineering, RV College of Engineering, Bangalore, Karnataka India.
New Gener Comput. 2023;41(2):213-224. doi: 10.1007/s00354-023-00212-7. Epub 2023 Mar 13.
World Health Organization (WHO) proclaimed the Corona virus (COVID-19) as a pandemic, since it contaminated billions of individuals and killed lakhs. The spread along with the severity of the disease plays a key role in early detection and classification to reduce the rapid spread as the variants are changing. COVID-19 could be categorized as a pneumonia infection. Bacterial pneumonia, fungal pneumonia, viral pneumonia, etc., are the classifications of several forms of pneumonia, which are subcategorized into more than 20 forms and COVID-19 will come under viral pneumonia. The wrong prediction of any of these can mislead humans into improper treatment, which leads to a matter of life. From the radiograph that is X-ray images, diagnosis of all these forms can be possible. For detecting these disease classes, the proposed method will employ a deep learning (DL) technique. Early detection of the COVID-19 is possible with this model; hence, the spread of the disease is minimized by isolating the patients. For execution, a graphical user interface (GUI) provides more flexibility. The proposed model, which is a GUI approach, is trained with 21 types of pneumonia radiographs by a convolutional neural network (CNN) trained on Image Net and adjusts them to act as feature extractors for the Radiograph images. Next, the CNNs are combined with united AI strategies. For the classification of COVID-19 detection, several approaches are proposed in which those approaches are concerned with COVID-19, pneumonia, and healthy patients only. In classifying more than 20 types of pneumonia infections, the proposed model attained an accuracy of 92%. Likewise, COVID-19 images are effectively distinguished from the other pneumonia images of radiographs.
世界卫生组织（WHO）宣布冠状病毒（COVID-19）为大流行病，因为它感染了数十亿人并导致数十万人死亡。随着疾病变体的变化，疾病的传播及其严重程度在早期检测和分类中起着关键作用，以减少其快速传播。COVID-19可归类为肺炎感染。细菌性肺炎、真菌性肺炎、病毒性肺炎等是几种肺炎形式的分类，这些又细分为20多种形式，COVID-19属于病毒性肺炎。对其中任何一种的错误预测都可能误导人们进行不当治疗，从而危及生命。通过X光图像等射线照片可以对所有这些形式的肺炎进行诊断。为了检测这些疾病类别，所提出的方法将采用深度学习（DL）技术。使用该模型可以实现对COVID-19的早期检测；因此，通过隔离患者可以最大限度地减少疾病的传播。为了便于执行，图形用户界面（GUI）提供了更大的灵活性。所提出的模型是一种GUI方法，它通过在Image Net上训练的卷积神经网络（CNN）使用21种肺炎射线照片进行训练，并将它们调整为射线照片图像的特征提取器。接下来，将这些CNN与联合人工智能策略相结合。对于COVID-19检测的分类，提出了几种方法，这些方法仅涉及COVID-19、肺炎和健康患者。在对20多种肺炎感染进行分类时，所提出的模型达到了92%的准确率。同样，COVID-19图像也能有效地与射线照片中的其他肺炎图像区分开来。
