In this project, I focused on leveraging advanced medical imaging techniques to enhance the detection of malaria. The goal was to improve the accuracy and efficiency of the diagnostic process using deep learning models. Here’s a breakdown of the key steps involved:
Key Steps:
Image Quality Enhancement: To ensure the accuracy of our models, I implemented various techniques to improve image quality. This included noise reduction to eliminate irrelevant disturbances and contrast enhancement to highlight important features in the medical images. These preprocessing steps were crucial for enhancing the performance of the subsequent model.
Data Augmentation: To address issues of data scarcity and variability, I employed advanced data augmentation techniques such as CutMix and Mixup. I make comparison of these data augmentation techniques to check their suitability for our use case and selected the Mixup data augmentation technique based on model performance. These methods help in creating synthetic variations of the training images, thereby improving the robustness and generalization of the model. The paper of CutMix and Mixup data augmentation is given below for more context:
CutMix Data Augmentation: Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., & Yoo, Y. (2019). Cutmix: Regularization strategy to train strong classifiers with localizable features. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 6023-6032).
Mixup Data Augmentation: Zhang, H., Cisse, M., Dauphin, Y. N., & Lopez-Paz, D. (2017). mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412.
Model Training and Hyperparameter Tuning: I closely monitored the training process of our deep learning model using Weights & Biases (WandB). This tool allowed me to track various metrics and adjust hyperparameters to optimize the model’s performance, ensuring it learned effectively from the data.
Model Deployment: The final step involved deploying a Convolutional Neural Network (CNN) for the classification task. The CNN was trained to distinguish between malaria and non-malaria cases based on the medical images. This deep learning model proved to be effective in automating the diagnostic process, providing accurate and timely results.
