| Name | Slack Name |
|---|---|
| Vaibhav Patel | @Vebby |
| Shudipto Trafder | @Shudipto Trafder |
| Sankalp Dayal | @SankalpDayal |
We have developed a model that can detect pneumonia from Chest X-Rays of the patient which has a significant level of accuracy in detecting pneumonia in comparison with practicing radiologists. Detecting pneumonia from chest radiograph is a tough task for the radiologist. The appearance of pneumonia in X-ray images are often confusing, can overlap with other diagnoses, and can mimic many other abnormalities. So the radiologists can get confused by this, leading to waste their time as well as energy just to detect the disease like pneumonia from the radiograph. So to help them get a second opinion, they can take help of our model for the detection of pneumonia.
For this project, we have dataset present on kaggle.The dataset is organized into 3 folders (train, test, val) and contains subfolders for each image category (Pneumonia/Normal). There are 5,863 X-Ray images (JPEG) and 2 categories (Pneumonia/Normal).
Chest X-ray images (anterior-posterior) were selected from retrospective cohorts of pediatric patients of one to five years old from Guangzhou Women and Children’s Medical Center, Guangzhou. All chest X-ray imaging was performed as part of patients’ routine clinical care.
For the analysis of chest x-ray images, all chest radiographs were initially screened for quality control by removing all low quality or unreadable scans. The diagnoses for the images were then graded by two expert physicians before being cleared for training the AI system. In order to account for any grading errors, the evaluation set was also checked by a third expert. A glimpse of Dataset.
Dataset can be found on this site-Pneumonia Dataset
Data: https://data.mendeley.com/datasets/rscbjbr9sj/2
License: CC BY 4.0
Citation: http://www.cell.com/cell/fulltext/S0092-8674(18)30154-5 Inspiration Automated methods to detect and classify human diseases from medical images.
Prerequisite packages which should be there in your run environment to run this project model.
| Package Name | Version Number |
|---|---|
| Torch | 1.2.0 |
| Torch vision | 0.4.0a0+6b959ee |
| Numpy | 1.17.0 |
| Matplotlib | 3.0.3 |
| PIL | 5.4.0 |
We have tried various other network model for this project. Below shown table shows the comparision of various models and their effect of the accuracy of prediction.
| Model Name | Total Accuracy |
|---|---|
| VGG 16 | 79.2188 |
| RESNET 50 | 83.2812 |
| RESNET 101 | 91.7188 |
Paper on RESNET by arvix We used transfer learning Method And used Resnet101 pretrained model
We used transfer learning, and replace the FC layer of RESNET101 with below layers
| Parameters | Description/Status |
|---|---|
| Linear layer | 3 Layers |
| Dropout | 0.4-0.5 |
| Activation Function | Mila |
| Last Layer Activation Function | LogSoftmax |
| Parameters | Description/Status |
|---|---|
| Batch Size | 32 |
| Epoch | 10 |
| Freeze Status | Unfreeze |
| Learning Rate | 0.0001 |
| Optimizer | Rectified Adam (RAdam) |
| Loss Function | Cross Entropy Loss |
FLOW DIAGRAM
Total Accuracy: 91.7188 %
Class wise accuracy:
| Class Name | Accuracy |
|---|---|
| NORMAL | 79% |
| PNEUMONIA | 98% |
- To test this model on NIH Chest X-ray Dataset to study the performance of this model.NIH Chest X-ray Dataset
- Comparison of the accuracy of detecting pneumonia between the proposed model and by radiologists.
- To make an Andriod and IOS application after successfully acheiving state of the art accuracy.