• Overview
  1. Classification of Pneumonia
     • Infection that occurs in the parenchymal tissue s of the lungs
     • Classified into viral and bacterial pneumonia according to the causative organism
     • Chest X-ray is commonly used as screening tools due to their low-cost, easy accessibility, and fast imaging acquisition
  2. Model description
     • Input: Chest X-ray image
     • Output: Pneumonia or Normal (binary classification)
     • Model Architecture: ResNet-50 with attention mechanism (CBAM: Convolutional Block Attention Module)
  3. Performance
     • Quantitative evaluation
     • Qualitative evaluation
     • (a) original image, (b) normalized image, (c) Grad-CAM overlay of ResNet-50 feature map, and (d) Grad-CAM overlay of ResNet-50 with CBAM feature map
• Limitations
  • This model cannot classify the types of pneumonia (viral, bacterial).
  • It is difficult to classify using images containing areas other than the chest. (In Future work, we will add modules to crop only the lung area)
  • CT, MRI images are not available

• Trade-offs
  • Depending on the resolution of the X-ray image you have, normalization method used for preprocessing may change.
  • EX) z-norm normalizaiton, min-max normalization, histogram equalization, etc.

• Performance
  • Quantitative evaluation: Accuracy, Sensitivity, Specificity, PPV, NPV
  • Qualitative evaluation: Grad-CAM visualization

• Test your own data
  • See how the model works on your own image here (we will not keep a copy) !
  • This function will be added later.

• Provide feedback
  • We’d love your feedback on the information presented in this card and/or the framework we’re exploring here for model reporting.
  • Please also share any unexpected results. ♥

참고: https://modelcards.withgoogle.com/face-detection#overview