Inspiration

Chest X-rays are widely recognized as an effective tool to screen, diagnose, and manage a variety of other thoracic diseases, including COVID-19, and the World Health Organization (WHO) identifies radiography as the best available method to diagnose pneumonia.7 The challenge is that billions of people worldwide, particularly those in rural and remote regions, do not have timely access to radiologists who can interpret the results of chest scans.

What it does

Accrad’s solution is designed to help detect 15 thoracic diseases, including two of the most deadly: COVID-19 and viral pneumonia. Pneumonia is the largest cause of death in children worldwide, with overall mortality highest in sub-Saharan Africa, and COVID-19 caused over 1.5 million deaths worldwide in 2020 alone.5,6 The exponential rise in infections during the COVID-19 pandemic highlights the importance of early detection to curb infection and mortality rates.

How we built it

Accrad, a medical AI software company based in South Africa, has developed one such AI-powered solution, which is called CheXRad. Accrad’s internal testing found that CheXRad is capable of labeling certain pathologies in chest radiographs up to 160x faster than radiologists, at comparable levels of accuracy, sensitivity, and specificity.4 The Windows-based CheXRad application is preconfigured with a COVID-19 and viral pneumonia classification neural network trained using the Intel® AI Analytics Toolkit on Intel® DevCloud for oneAPI. It is optimized for deployment with the Intel® Distribution of OpenVINO™ toolkit on Intel® DevCloud for the Edge. To create new machine learning models or refine an existing model, healthcare researchers can quickly retrain and optimize using the Intel® Optimization for TensorFlow, available as part of Intel AI Analytics Toolkit. In addition, a desktop application was developed using Intel Distribution of OpenVINO toolkit for easy deployment, so radiologists could evaluate large numbers of image results. CheXRad thus offers a powerful, flexible, and scalable solution for healthcare researchers and for system integrators with Intel® processor-powered systems that support the Windows 10 IoT Edge operating system.

Challenges we ran into

The WHO estimates that two-thirds of the world’s population lacks access to basic radiology services like X-rays.3 Kenya has only about 200 radiologists to serve a population of 43 million people,8 and many other countries—including Liberia, South Africa, Australia, the United States, and the United Kingdom—have identified similar shortages in the number of available radiologists. Even before the COVID-19 pandemic, people in many parts of the world sometimes had to wait weeks for a radiologist to read their chest scans, leading to worsened patient outcomes and increased spread of infectious diseases.9,10 AI-powered solutions are designed to help radiologists manage their increasing workloads through a variety of methods, including by screening large volumes of radiographs to predict those with a high probability of disease.

Accomplishments that we're proud of

Radiologists, physicians, and researchers can use CheXRad to help identify COVID-19, viral pneumonia, and other diseases on chest X-ray images and predict which patients are most likely to need a ventilator or medication. The solution provides the probability value of each pathology, as well as specifying regions of interest for radiologists to pursue. In internal testing, Accrad found that CheXRad could label certain pathologies in 140 chest radiographs in just 90 seconds—up to 160x faster than radiologists, at comparable levels of accuracy, sensitivity, and specificity. CheXRad has been deployed using Microsoft Partner Center and will be available on the Microsoft Store. The application is designed for use in the healthcare research field, including in radiology labs, clinics, and hospitals, and it can be used by systems integrators with Intel processor-based systems that support the Windows 10 IoT Edge operating system

What we learned

The WHO estimates that two-thirds of the world’s population lacks access to basic radiology services like X-rays.3 Kenya has only about 200 radiologists to serve a population of 43 million people,8 and many other countries—including Liberia, South Africa, Australia, the United States, and the United Kingdom—have identified similar shortages in the number of available radiologists. Even before the COVID-19 pandemic, people in many parts of the world sometimes had to wait weeks for a radiologist to read their chest scans, leading to worsened patient outcomes and increased spread of infectious diseases

What's next for Accrad

Accrad is building Africa’s largest medical AI software company. The company is devoted to conquering cancer and other pathologies across Africa and is developing AI software to help radiologists save time and diagnose with accuracy under pressure.

Share this project:

Updates