This repository provides a common framework for evaluating the quality of lung image datasets. The tool performs multiple assessments and generates performance metrics, including classification, unsupervised Region of Interest (ROI) generation, and segmentation.
For an indepth understanding of the code, please refer to :
Rajasekar, Elakkiya, Harshiv Chandra, Nick Pears, Subramaniyaswamy Vairavasundaram,
and Ketan Kotecha. "Lung image quality assessment and diagnosis using generative
autoencoders in unsupervised ensemble learning." Biomedical Signal Processing
and Control 102 (2025): 107268.
-
Classifier: A Deep CNN model to predict disease from input images.
Metrics: F1 Score, Classification Matrix, Precision, Recall, Accuracy, AUC-ROC. -
Segmentation: Evaluates segmentation quality using an ensemble model (U-Net, U-Net++, Segnet, FCN, NASNet).
Metrics: F1 Score, Precision, Recall, DICE. -
Unsupervised ROI Generation: Standardised segmentation generator for datasets to understand segmentation performance.
- Clone the repository:
git clone https://github.com/harshivchandra/LungDataQualityAssessment.git- Navigate to the git directory and run Jupyter Notebook:
cd directory_github/LungDataQualityAssessment
jupyter notebook EnsembleLearning.ipynb