Overview / Usage

The project harnesses the power of Deep Learning and Intel oneAPI to accurately diagnose pneumonia from X-ray images. By automating the detection process, we aim to expedite diagnosis, enabling timely medical intervention and improving patient outcomes. This solution can be seamlessly integrated into clinical workflows, enhancing efficiency and aiding healthcare professionals in making informed decisions. Whether in hospitals or remote areas, our model provides a reliable tool for pneumonia detection, ensuring widespread accessibility to quality healthcare.

Methodology / Approach

Our methodology revolves around employing Deep Learning techniques to tackle the challenge of pneumonia detection in X-ray images. Leveraging the** TensorFlow and Keras frameworks**, we utilize transfer learning with the VGG16 architecture for feature extraction. Additionally, we incorporate Intel's optimized libraries, such as scikit-learn-intelex for accelerated computations and oneDNN for deep learning optimizations, ensuring efficient execution on Intel DevCloud. Our model is trained on a large dataset, augmented using ImageDataGenerator for robustness. With a focus on accuracy and efficiency, our approach enables swift and reliable diagnosis of pneumonia, contributing to improved healthcare outcomes.

Technologies Used

Intel Technologies:

- Intel Optimization for TensorFlow (utilized through scikit-learn-intelex)
- Intel oneDNN (oneAPI Deep Neural Network Library)
- Intel Math Kernel Library (mkl)
- Intel DevCloud
- Intel Distributions for Python

Libraries and Frameworks:

- TensorFlow
- Keras
- scikit-learn-intelex
- numba
- scipy
- matplotlib

Tools:

- Intel DevCloud
- Kaggle 

​

Repository

https://github.com/reeshmashahiras/chest-x-ray-pneumonia-detection

Other links

• Dataset