Implemented a Deep CNN model for Detection of pneumoconioses Patients Using Chest X-Ray Images. Built the model using Transfer Learning by fine-tuning pre-trained model DenseNet121 that have been pre-trained on ImageNet dataset.
Fifty-five Chest X-Ray diagnosed with pneumoconioses from B-reader Syllabus and sixty-one normal Chest X-Ray from CheXPert were used in the project. Ninty-two Chest X-rays (46 with pneumoconioses and 46 normal Chest X-rays) were used train the model and 24 Chest X-rays (12 with pneumoconioses and 12 normal Chest X-rays) were used as test set.
The Chest X-rays were enhanced using Bi-Histogram Equalization with Adaptive Sigmoid Functions algorithm (BEASF) and Contrast Limiting Adaptive Histogram Equalization (CLAHE).
The confusion matrix of the test set is:
The history of loss and accuracy plot for 50 epochs is:
The classfication report is:
