Tuberculosis (TB) caused 1.7 million deaths in 2016, surpassing the number of HIV/AIDS deaths globally. In that year, 10.4 million people fell ill with TB, mostly from poor and vulnerable communities. As a result of these troubling statistics, the World Health Organization (WHO) has accelerated its fight against TB with its End TB Strategy.
The WHO estimates that 3.6 million people with TB are missed by health systems every year and do not receive adequate care. This is primarily because patients with TB may present with mild or no symptoms, particularly early on. Many affected people arrive at clinics too late with advanced disease or multi-drug resistant TB (MDR-TB), which is difficult to treat and more likely to cause death.
One of the key strategies to tackling TB is early diagnosis. Unfortunately, sputum testing is only 50% accurate and frequently misses the disease in its early stages. Molecular testing – while highly accurate – is too expensive for population screening in most regions.
A Promising Solution for Early Detection
One promising solution for early detection of TB involves chest radiography. In particular, the cost of radiology equipment, namely digital radiography machines, has substantially decreased over the past five years, increasing the availability of the technology worldwide. However, the adoption and utilization of this technology has lagged in the developing world, as there is a lack of trained radiologists and medical expertise to interpret images.
More recently, there has been interest in using artificial intelligence (AI) for medical image interpretation. In 2017, Dr. Paras Lakhani published work on using AI for TB detection1 on chest radiographs and was able to achieve detection accuracy greater than 95%. The results were clinically validated at Jefferson Radiology. Since then, Dr. Lakhani and SemanticMD, a company specializing in AI solutions for medical imaging, partnered to commercialize this screening solution. The solution has the potential to lead to earlier detection and make a tremendous impact in battling TB.
How it works
Deep learning technologies enable the accurate and rapid detection of TB. SemanticMD’s algorithm is trained on a growing database of thousands of images from the U.S., India, China, Eastern Europe and South Africa. Based on this analysis a score for each patient is computed. The software can run automatically after a digital X-ray is captured. It supports DICOM and HL7 standards to route images which enable integration with hospital IT infrastructure. In addition, the AI solutions can be deployed offline with a low power device which is crucial for remote, low-resource settings.
|
|
|
SemanticMD aims to scale its AI solution to provide low-cost, accessible TB detection to vulnerable populations, particularly throughout Southeast Asia, China and Africa. The company is already working with partners in China, South Africa, The Gambia, Rwanda and Nigeria – offering instant detection for less than $1 per scan. The solution is scalable, with the ability to integrate with devices from any X-ray manufacturer. It can be accessed from the cloud or deployed locally (where internet access is limited).
The SemanticMD algorithm development platform is the engine behind its TB solution. With the ability to rapidly develop algorithms, the company is tackling other health screening concerns including the detection of breast cancer, lung cancer, malaria and diabetic retinopathy.
With focused investment to scale AI solutions for automated disease detection, millions of currently underserved people can gain access to high quality, affordable health screening.
Learn more: https://semantic.md/ai-tb.html
References
- Lakhani P, Sundaram B. Deep Learning at Chest Radiography: Automated Classification of Pulmonary Tuberculosis by Using Convolutional Neural Networks. Radiology. 2017; 284(2):574-582.