How Real is the Power of Artificial Intelligence in Diagnostics and Clinical Medicine?
AI is here and it’s here to stay. There is no doubt that AI-enabled solutions are redrawing the healthcare landscape and are inviting apprehensions of them taking over human jobs. However, such fears may be misdirected as these cognitive systems are more meant to augment professional judgment rather than replace it.
Being a cardiologist, an entrepreneur founding Medwin hospital, and now a venture capitalist investing in digital health and medical devices start-ups, I have been in the midst of several developments and changes in the healthcare industry. The Indian healthcare sector, in particular, has been full of opportunities and challenges are driven by factors like the increased prevalence of lifestyle diseases, increased health awareness, rising income levels, and aging population. The fast-growing trend of telemedicine and the role played by smartphones and wearables in monitoring the health of individuals are game-changers in the way healthcare services will be delivered. However, India is still beset by numerous obstacles like the health infrastructure, access to quality healthcare in rural areas, scarcity of trained medical emergency personnel and lack of focus on preventive care. At Endiya Partners, we believe the answer to many of India’s healthcare woes lie in the breakthrough innovations by entrepreneurs of the country with, technologies like AI being the key enabler.
Today, we are witnessing the proliferation of AI/ML technologies in every field. From autonomous cars to intelligent behavioral analytics, technology giants, as well as start-ups, are innovating to leverage the power of AI to create better solutions. Healthcare is an obvious domain, where the role of AI has tremendous implications. AI has already made advances in several areas like wearable technology for patients, clinical decision making, drug discovery, hospital administrative management, chatbots, virtual assistants and robotic surgery.
However, in the Healthcare AI space, the area that is attracting the most investments in Medical Imaging and Diagnostics; and for good reason. Sifting through large amounts of imaging data, lab values, pathology reports and information captured in EMRs poses a huge challenge to physicians in terms of drawing actionable insights. Machine Learning tools, on the other hand, can eliminate the time-consuming tasks of interpreting large data sets and are not prone to the fatigue and biases that humans have.
When we look at the Indian scenario, the ratio of doctors to citizens is a mere 1: 1700. This puts a lot of pressure on doctors to diagnose and treat a large number of patients on a daily basis. AI-enabled technology allows for rapid and accurate diagnosis that has the potential to completely re-define this space.
Let’s dive deeper into the advancement of AI in the fields of Radiology, Pathology, Ophthalmology, and Dermatology
Machines armed with deep learning algorithms are capable of recognizing visual patterns that help Radiologists get a better understanding of the patient’s condition. Recently, Stanford researchers, led by Andrew Ng, adjunct Professor of Computer Science, developed an algorithm that offers diagnoses based on chest X-ray images. They did this by training the algorithm to learn from hundreds of thousands of chest X-rays to diagnose up to 14 types of medical conditions including pneumonia which it diagnosed better than expert Radiologists working alone.
The Massachusetts General Hospital (MGH), is already using AI to complement the work of radiologists. In 2016, MGH partnered with global computing technology company NVIDIA to develop deep learning algorithms to program a server designed for AI applications and used MGH’s phenotypic, genetic, and approximately 10 billion images to train it. Although more than a dozen other algorithms are still in the works, MGH Paediatric Radiologists currently use one to estimate bone age. After exposing the algorithm to thousands of images and teaching it to accurately assess bone age, radiologists now use the tool to determine a patient’s bone age based on X-rays.
While there are unanswered questions over the adoption of AI including regulatory challenges, ethical challenges, and patient acceptance, the research and trends do seem to suggest that Radiology is one area that will be revolutionized by AI in the long term.
When it comes to Pathology, it is the most definitive way of diagnosing disease and influences a vast majority of the decisions made. That also means that it has potentially the largest probability of misdiagnosis. So, when we bring AI-powered technology into the process, the ability of machines to recognize complex patterns enables pathologists to gain more insights than what a human eye can detect.
In a study carried out by researchers taking part in Google’s Brain Residency Program, an algorithm was trained to detect breast cancer tumors in a dataset of digitized pathology slides provided by a Dutch medical institute. After ‘training’ the algorithm, researchers were able to achieve a 92% sensitivity in picking out tumor cells from the slides – significantly higher than the 73% achieved by trained Pathologists with no time constraint.
One of our portfolio companies SigTuple is a start-up that uses computer vision and AI for diagnosis. It does this by first digitizing the slides of blood samples as well as other samples and then using its AI engine to classify and tag various elements in the images. Their core product Manthana is a continuous learning platform that is used to churn data to generate intelligence and integrates the capabilities and workflows of SigTuple.
In India with urban diagnostic facilities being crowded and the rural population being underserved, applications like those provided by SigTuple can pave the way for improved accuracy and speed of medical diagnostics and a much-needed change.
Today, intelligent machine algorithms can detect conditions like Diabetic Retinopathy, which is something only experienced specialists can do. Given the acute shortage of qualified ophthalmologists, AI has the potential to fill the gap in this field. Google’s DeepMind, a leader in AI research and its applications in healthcare, can analyze OCT scans using deep learning to identify a retinal disease. Using software engineered from Google, researchers at Stanford trained computer models to get better at interpreting photographs of the fundus of the eye, showing that their model can successfully pick out an image of an eye damaged by Diabetic Retinopathy
With developments like these, experts are confident that Ophthalmology will be one of the first branches of medicine to be fundamentally transformed by AI.
Dermatology is another field in medicine where Al has made tremendous progress. At the University of California San Francisco (UCSF), a proposal for an artificial intelligence-based skin cancer screening tool won the 2017 Cancer Center, Impact Grant. The winning team is developing a scanning tool for detecting melanoma, which has a 100 percent survival rate if found early.
Other types of innovations are AI-driven mobile apps to diagnose common skin conditions such as pimples, acne, and scars and recommend treatment regimens. Given the low ratio of Dermatologists to people in India (1:100,000) and the majority of them being in the metro cities, there is vast potential for applications like these to enable effective teleconsultations.
It’s not all hunky-dory
While machine-learning promises to dramatically improve the efficiency and effectiveness of medical diagnosis and healthcare in general, it is not without its limitations. As these algorithms grow smarter, it adds an additional layer of complexity and opaqueness concerning machine behavior, that give out recommendations without explanation or insight. This lack of transparency becomes a concern in a critical field like healthcare.
Also, one cannot ignore the pitfalls of a machine-only approach. There is no substitute for a skilled physician’s expertise in considering the unique clinical situation of each patient. Hence, the best combination is where the physician incorporates these AI-enabled tools in their practice, to reduce the menial tasks and enhance their ability in effective decision making.
The present and not so distant future
AI is here and it’s here to stay. There is no doubt that AI-enabled solutions are redrawing the healthcare landscape and are inviting apprehensions of them taking over human jobs. However, such fears may be misdirected as these cognitive systems are more meant to augment professional judgment rather than replace it. From the ability of chatbots to answer diagnostic questions on minor issues to complex machines and algorithms detecting cancer, both patients and physicians stand to benefit from the adoption of AI. If we overcome the impending challenges that come with incorporating new technology in the healthcare process, it will probably have the most disruptive and beneficial impact yet in the medical profession.
At Endiya, we continue to explore start-ups in the deep tech healthcare space that will bring unsurpassed value to treatment as well as prevention, and we look forward to working with them in becoming pioneers of the industry.