Xilinx teams up with AWS and Spline.AI to develop deep learning models for X-rays

Xilinx, AWS, and Spline.AI have collaborated to develop a deep learning model for medical X-ray and disease detection that can be used in clinical medicine settings for pneumonia and COVID-19 prediction.

The research team relied on 30,000 carefully selected and labeled pneumonia images and 500 COVID-19 images to train deep learning models to improve the accuracy and speed of predictions. The training data will be made available to public research and medical institutions including the National Institutes of Health (NIH), Stanford University and the Massachusetts Institute of Technology, Xilinx said in a statement on Tuesday.

Functional deep learning models are available along with a reference design kit priced at $1,295. According to the description, the Xilinx Zynq UltraScale+ MPSoC is added to the ZCU104 FPGA development board provided as an edge device. The Xilinx deep learning processor unit is integrated into the MPSoC (multi-processor system-on-chip), which accelerates convolutional neural network processing in AWS IoT Greengrass.

Python-PYNQ is an open-source programming platform that allows clinical researchers to adapt the board for different applications. Models can be developed in mobile, portable, or point-of-care situations and are able to scale using the cloud. AI models are trained using Amazon SageMaker and deployed from the cloud to the edge via AWS-IoT-Greengrass, allowing remote learning model updates.

AWS said the combined technologies mean that highly accurate clinical diagnoses can be made using low-cost medical devices. Dirk Didascalou, vice president of IoT at AWS, said doctors could upload X-ray images to the cloud without the need for physical medical equipment, allowing doctors to perform remote diagnoses or treatments.

After validation and regulatory approval, the system can be deployed for clinical use. “It depends on the OME of the medical device or the deployment in the clinic or hospital,” said Subh Bhattacharya, director of medical, medical devices and science at Xilinx. “The reference design kit can be used to accelerate time to market to develop and deploy other models with different radiology or clinical processes.”