1st International and 10th National Iranian Conference on Bioinformatics

Home Accepted Papers

Diagnosis of COVID-19 patients based on chest CT images using image processing algorithms

Paper ID : 1272-ICB10

Authors:

Ahmadreza Rasouli *¹, محسن خدارحمی², جواد ظهیری³

¹دانشجوی دکترای تخصصی بیوانفورماتیک دانشگاه تهران

²دپارتمان رادیولوژی بیمارستان شهید مدنی کرج

³محقق پسا دکترا دانشگاه سندیگو کلفرنیا

Abstract:

The rapid diagnosis of the disease often makes the treatment procedure faster and less expensive. In the case of SARS-COV2, an infectious disease with a high rate if transmission, diagnosis without the need to see a doctor is of great importance.[1]
In this study, we aim to introduce an algorithm with a high performamce based on which we can diagnose a patient based on the chest CT imaging features, and without any need for the suspected patient to see a doctor.
Firstly, we enhanced the quality of CT images by using the classic algorithms and the most important filters for image processing.[2] Then thr most important CT imaging feature were extracted using a convolutional neural network. We have used two imaging sets including 42 images from patients and 42 images from healthy persons. For each of CT features we assigned a certain weight.
Finally, we desgined an algorithm which gets a CT image from an individual as input, and determines whether this individual is healthy or patient, by enhancing the quality of the initial image, extracting the relevant imaging features by using a convolutional neural network, and adding the multiplication of each feature and its associated weight.
In this study, by examining 1288 photos of healthy people and 1343 photos of sick people, we reached about 90% accuracy in diagnosing the disease.

Keywords:

Corona-Image processing-Neural Network-Feature extraction-Preprocessing

Status : Paper Accepted (Poster Presentation)