Disease Detection using AI Reference Kits from Intel: Developer Spotlight

by

Ugonna Chikezie, Software Technical Marketing Engineer

Intel Corporation

Dev Aryan Khanna is an Intel® Student Ambassador for oneAPI with a research focus on AI/ML. He has recently worked on the Healthcare AI Companion project, an intelligent healthcare tool combining advanced deep learning models for image classification and text-based symptom analysis. The tool helps with the early detection of diseases based on X-ray images and symptom descriptions. It leverages AI Reference Kits from Intel  and oneAPI libraries for model optimization.

About the Healthcare AI Companion Project

The motivation to solve the challenge of creating the Healthcare AI Companion stemmed from the inherent limitation of existing healthcare solutions optimized solely for specific GPUs. This posed a significant problem as it hindered adaptability to Intel® CPUs, impacting the global client base. The project aimed to address this gap by developing an intelligent healthcare tool optimized for diverse architectures, specifically focusing on gaining efficient performance on Intel® hardware.

The approach to building the Healthcare AI Companion involved leveraging cross-architecture oneAPI tools. These tools empower the development of single-language and platform applications that can seamlessly be ported to and optimized for various architectures, including Intel’s CPUs. These tools ensure versatility, efficiency, and accessibility, aligning with the project's goal of providing a comprehensive and adaptable healthcare solution.

Intel® Hardware and Software Optimizations Used

1. Intel® oneAPI Deep Neural Network (oneDNN) Library

• How: Graphs in Fig.1 illustrate the accelerated performance of CNN layers achieved with oneDNN for image-based disease detection.
• What: The library optimizes convolutional layers, crucial for enhancing the efficiency of disease detection in X-ray images.
• When: Throughout the project life cycle, particularly during the training and inference stages.

Fig 1: Improved Performance Results with oneDNN

2. Intel® oneAPI Data Analytics Library (oneDAL)  

• How: Fig.2 shows the expedited feature engineering tasks of the data preprocessing pipeline performed using oneDAL..
• What: While not the primary focus, oneDAL contributes to ensuring data quality and enhancing feature engineering, crucial for accurate disease detection models.
• When: During the data preparation phase before model training.

Fig 2: Accelerated Feature Engineering using oneDAL

3. Intel® Extension for PyTorch*

• How: The comparative chart in Fig.3 demonstrates the acceleration achieved in model training with Intel Extension for PyTorch. The PyTorch optimizations also enable mixed-precision training.
• What: Intel Extension for PyTorch optimizes model training, accelerates deep learning workloads on Intel CPUs, and enables mixed-precision training without sacrificing accuracy.
• When: Primarily during the model training phase.

Fig.3: Accelerated model training with Intel Extension for PyTorch

Other Intel® Software Tools and Libraries:

• Intel® Extension for TensorFlow*
• Intel® Integrated Performance Primitives (Intel® IPP)

AI Reference Kits from Intel:

• Disease Prediction Using Natural Language Processing
• Medical Imaging Diagnostics Using Computer Vision
• Image Data Generation

Intel® Hardware:

• 4th Gen Intel® Xeon® Scalable Processor on the Intel® Developer Cloud platform.

Success Story of the Intel Student Ambassador

Dev Aryan Kanna is an undergraduate student at the Guru Gobind Singh Indraprastha University in India. As an Intel Student Ambassador for oneAPI, his journey has been enriching. Hosting workshops, hackathons, and working on projects as parts of the Intel Student Ambassador Program have provided valuable hands-on experience to the student attendees and enhanced his own learnings. Access to abundant resources, including the latest Intel® hardware through the program, has deepened his understanding of Intel® technologies, fostering a passion for innovative solutions.

The Healthcare AI Companion project, which secured the 1st runner-up position in a recent Intel Student Ambassador Hackathon, has fueled his goal to leverage technology for healthcare advancements. With ongoing access to Intel's resources, he aims to further contribute to the intersection of AI and healthcare for impactful solutions.

Useful Resources

• Intel oneAPI Programming Model
• Intel’s AI Software Portfolio
• Intel’s AI/ML Frameworks Optimizations
• AI Reference Kits from Intel
• Intel Developer Cloud
• Intel Student Ambassador Programs

Notes and Disclaimers

Performance varies by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex​.  

Your costs and results may vary. 

Intel technologies may require enabled hardware, software, or service activation.

Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.

© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. 

*Other names and brands may be claimed as the property of others.  ​