Real-time pneumonia prediction using pipelined spark and high-performance computing.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Pneumonia diagnosis relies heavily on chest X-rays, but analysis is complex and time-consuming.
• Rapidly growing medical data necessitates efficient diagnostic models.
• Existing methods for pneumonia prediction lack precision and speed.

Purpose of the Study:

• To develop a faster and more accurate system for pneumonia diagnosis using X-ray images.
• To leverage distributed deep learning for accelerated model training.
• To reduce computational resources required for pneumonia prediction.

Main Methods:

• Utilized a distributed data-parallel approach with high-performance computing for model training.
• Implemented a deep learning model deployed in Spark for scalability and acceleration.
• Employed concurrent training across multiple compute nodes to enhance efficiency.

Main Results:

• Achieved a prediction speed 1.5 times faster than traditional Convolutional Neural Network (CNN) models.
• Attained a high accuracy of 98.72% in pneumonia prediction.
• Demonstrated speed-ups ranging from 1.2 to 1.5 in parallel models.

Conclusions:

• Distributed deep learning significantly reduces training time for pneumonia detection models.
• The proposed method offers a precise and efficient solution for early pneumonia diagnosis.
• This approach is vital for handling large datasets and enabling timely medical interventions.