Osteoarthritis — the wearing down of the protective cartilage between the bones — causes severe joint pain and is a leading cause of disability. Yet this common disease of the bones is one of the most difficult to detect and prevent in its early stages.
In the U.S. alone, osteoarthritis is responsible for the majority of total knee and hip replacements. The most prevalent form of the disease affects the knees, occurring in 10 percent of men and 13 percent of women over the age of 60. These numbers are only expected to grow due to aging populations and the obesity epidemic — with huge costs on public health systems and well-being.
Hoping to turn this trend around, ImageBiopsy Lab, an Austrian startup and member of our Inception program, is using deep learning to diagnose osteoarthritis of the knees much more efficiently and cost-effectively.
Better, Faster, Stronger
The progression of osteoarthritis is currently considered to be largely unstoppable. Diagnosis relies purely on the visual assessment of two-dimensional X-ray images by doctors, combined with an understanding of the patient’s background. This is a time- and resource-intensive process, which can prolong the wait for an accurate diagnosis and precise treatment plan.
No medications can cure osteoarthritis, so treatments focus on the relief of symptoms and restoring functionality. This means that the quicker and more accurate the initial diagnosis, the better osteoarthritis can be managed.
Using computer vision and deep learning algorithms, ImageBiopsy Lab is enabling doctors to gain a precise, three-dimensional understanding of two-dimensional images. The company trained its algorithms, using NVIDIA GPUs, on over 150,000 radiographs, so doctors can receive accurate measurements of the areas around the bones in the knees. The results can indicate the severity of the patient’s osteoarthritis without any further processing needed.
Instead of having to load the X-rays from the server and then manually assess the images, doctors receive an automated analysis of radiographs, performed in real time, indicating the strength of the bone. With these processes automated, doctors can enjoy time savings of up to 90 percent compared to manual image grading, giving them more time to focus on diagnosis and building treatment plans. Through automatizing the initial stages, doctors can now make complete diagnoses for four times as many patients in the amount of time it currently takes to diagnose one.
Plus, with objectively measured parameters in hand, doctors can initiate clearly defined treatment pathways, which can slow the progression of the disease.
“The power of NVIDIA’s GPUs has allowed us to accelerate the training of our networks and enabled us to deploy our product much faster,” said Richard Ljuhar, CEO and co-founder of ImageBiopsy Lab.
ImageBiopsy Lab is one of more than 2,200 startups in our Inception program. The virtual accelerator program provides startups with access to technology, expertise and marketing support.