Join a global team of 50 AI changemakers in this high-impact 2-month challenge to build AI solutions for reducing the risk of misdiagnosis in X-rays.
This challenge requires experience in Deep Learning and Machine Learning.
It is not unusual to miss radiological abnormalities. This is common because of areas on the chest X-ray where lesions can easily be overlooked: behind the clavicle, heart, diaphragm, at the hilum, and pleural lesions. Such mistakes lead to a diagnostic delay and significant delay in time to initiation of treatment and palliation of symptoms. Dr CADx’s AI solution helps doctors minimize the incidence of missed cases by picking up subtle findings at an early stage.
The deep learning algorithms of Dr CADx analyze medical images for features that suggest the presence of diseases. The insights brought by the analysis help the doctors reduce the risk of missed diagnosis by up to 20%.
Considering the initial target market in Africa, where each year, 45 million patients lack access to have a radiologist report on their medical image, Dr CADx’s solution has the potential to reduce the risk of misdiagnosis for at least 1.3 million patients annually (taking into account that at least 50% are chest X-rays and 30% are currently digital a proportion that is continuously growing).
Radiology is central to modern medicine, with the diagnosis of diseases frequently dependent on medical imaging (X-rays, CT scans, ultrasound, MRI), which needs to be interpreted by an imaging specialist, the radiologist. However, studies show that radiologists have limited accuracy, with up to 30% error rates. Additionally, due to the shortage of radiologists worldwide, it can take up to 30 days for about 10% of cases to be reported by radiologists in better-resourced regions like the EU. Furthermore, only 60% of medical images are read by radiologists.
The situation is even direr in regions like Africa, where radiologists read just about 10% of medical images. As a result, an estimated 45 million African patients get a medical image taken but do not have access to a radiologist to report on them. Given that the miss rates are higher among non-radiologists, reaching up to 45%, patient outcomes are non-ideal.
As a result of the 4.6 billion medical images taken worldwide annually, at least 1.4 billion are misdiagnosed. Patient outcomes can be severe, resulting in paralysis or death, with statistics showing that misdiagnoses result in 1.5 million deaths globally. Financially, errors in interpreting medical scans cost the global healthcare sector over $115 billion every year.
The Project goals
DR CADx has a model for 14 chest findings that have been already developed but not yet validated in a clinical study. Based on the initial market feedback from prospects, the following steps for the product development are thus as follows: Finetune the algorithm for the 14 chest findings to include:
1. Adding COVID, TB screening, and rib fractures, thus expanding it to 17 findings
2. Include more training data with a focus on data from Africa and different imaging equipment to further reduce bias and improve accuracy across various population groups
3. Improve the localization of detected findings
4. (Optional, if there is enough time) include analysis of lateral X-rays in addition to PA and AP images
5. (Optional, if there is enough time) a quality checking model to reject images that do not have chest X-rays
Why join? The uniqueness of Omdena AI Challenges
A collaborative experience you never had in your working life! For the next eight weeks, you will not only build AI solutions to make a real-world impact but also go through an entire data science project lifecycle. This covers problem scoping, data collection, preparation, and modeling for deployment.
And the best part is that you will join a global and collaborative team of changemakers. Omdena AI Challenges are not a competition or hackathon but a real-world project that will take your experience of what is possible through collaboration to a new level.