×
×

Customer challenge

COVID 19 diagnosis confirmation using polymerase chain reaction (PCR) is highly sensitive, time-consuming, and a complicated manual process that is in short supply

Scope

Develop an AI-based detection system to aid radiologists in rapidly and accurately interpreting radiography images to detect COVID-19 cases

Solution

  • Designed the architecture for COVID-19 detection using Convolutional Neural Network (CNN)
  • Prepared chest X-Ray images belonging to normal, Pneumonia, and COVID-19 infected persons
  • Trained the network, analyzed, and fine-tuned the network performance
  • The model generated after validation and testing, can classify the chest X-Ray images
  • The solution also shows heat map visualization of class activation maps for explainability

Benefits

  • Achieved higher detection accuracy (95%) compared to conventional image processing methods
  • Can assist the radiologist with real-time performance

Results

Detection of COVID-19 from X-Ray image demonstrator