What galvanized you to start at Qure.ai?
Being part of the Healthcare industry and interested in technology, I had spoken to dozens of startups in the last 4-5 years. All of them came with great ideas and strong engineering talent. Some of them had raised money from prominent investors in Silicon Valley. Very few of them had a clear purpose and a sustainable business model. I saw Qure.ai as a company with a clear purpose, the founders knew what they wanted to do and also what they did not want to do.
I was also impressed by the understanding and focus on the business aspects of AI deployment. They were asking important questions to get to the root of the customer use case, the value proposition, aspects of workflow and deployment, all key components of a clear business model. Overall, I also got a great rapport with the team members I met and sensed a good sense of humility among the great expertise they carried and that was what sealed it for me.
Tell us about your interaction with AI and other Intelligent Technologies that you work with, in your daily life.
I consider myself lucky to be living in an age where AI is helping us perform better. From determining the best route to reach work to AI-led bots that feed me the latest industry-specific updates, AI has completely transformed how we work. What I love is the fact that many of these AI engines are all invisible to the users whom we do not even realize that our search results are continuously improved by AI. That is my vision for how all AI should be, completely invisible and focused on getting the work done.
What is your vision about Qure.ai?
The key to reducing cost and improving access is early detection, something we all know and accept. For many clinical ailments, be it airborne diseases, accident induced trauma or even Cancer; early and accurate detection and makes a difference in both quality and quantity of survival. Early Detection and Diagnosis rely on the expertise of clinical professionals like Radiologists but while the volume of imaging exams is increasing exponentially, the number of radiologists being trained around the world is not keeping pace. In essence, there is a real and growing demand and supply gap that is impacting healthcare outcomes.
Our team sees the possibility to make a dent in this area. We are convinced that Machine Learning and image recognition technology can bridge some of these challenges and this is what brought about the inception of Qure.ai. Using AI, we aim to make quality healthcare affordable and accessible to everyone in society.
How do you plan to overcome the mistrust of AI usage in healthcare?
I’d start by saying that the mistrust is real and it should be there. Unlike a Navigation app wherein the worst result would be a delayed arrival, we are dealing with people’s health outcomes and therefore it is important for physicians to trust the tools they are using. That is one of the reasons why healthcare has been a highly regulated industry.
For Artificial Intelligence to be widely used in healthcare and for doctors and patients to benefit from it, it is essential that radiologists see the value in adopting it. For Radiologists to develop confidence in an algorithm it’s important that they understand why the algorithm has suggested a Diagnosis. This is one of the focus areas for our research team, and our algorithms are far from being a black box. They allow interpretability by letting the user see why and where the AI algorithm has detected an abnormality.
Another factor key to build is to ensure that the tools are generalizable and can be applied to diverse patient populations. In order to ensure this, Qure.ai has trained its algorithms on more than 7M datasets and works closely with some of the country’s leading radiologists to develop and fine-tune our products. We pride ourselves on having a larger number of the peer-reviewed research that adds to our credibility.
Lastly, I’d end with where I started. I could tell that all customers do not trust anyone, try out AI algorithms on your own data and see if it works for yourself. We offer all customers an option to try our algorithms before they deploy the same.
What differentiates Qure.ai ’s Deep Learning approach?
In imaging AI, most companies use a Deep Learning approach called “segmentation,” which requires radiologists to mark out abnormal regions on the radiology image (this process is called annotation). While it has some advantages, we believe that this marking out of abnormal regions is not scalable. For example, our X-ray database of 2.4 million X-rays might need about 160 years of annotation by a radiologist.
At Qure, we’ve done things somewhat differently. While we do use segmentation for certain Clinical areas, we also focused on another type of deep learning called, “classification,” which does not require abnormal regions to be marked. Instead, it works with a type of abnormality seen on a medical image. This type of labeling is a much simpler task.
In order to shorten the length of time it takes to do 2.4 million scans, we enabled automated labeling, by creating Natural Language Processing techniques that extract abnormalities, and their location, allowing us to label millions of scans in just a manner of minutes. In effect, our algorithms gain from the best of both worlds, the accuracy depth of Segmentation and the Coverage breadth plus speed of classification.
There are many other learnings that we have received and incorporated into our Deep Learning approach that I would like to keep confidential. And this is an ongoing journey, we are continuously seeking feedback from our users, learning from our deployments and also getting new and diverse datasets that make us improve our Deep Learning engines.
How can medical facilities, especially the ones in developing countries that do not have a fully-functional Electronic Health Record System, leverage on Qure.ai ’s products?
We’ve deployed solutions across many National and International locations. We are present in 14 countries, many of them are developing nations. We have 80 plus installations and 50,000 plus processed scans to date. As long as we can get a clear radiology image, our algorithms can be deployed in very frugal settings to do their job. We have proven this by deploying algorithms in screening vans in the Philippines and even remote clinics in Malawi, Africa or in small districts in India. This is another factor that differentiates us from our competitors. We understand that not every person in the developing countries has a strong IT department, some of them do not even have good internet access and we have responded by making our deployment approach flexible to meet the environments our customers operate in.
What are the core features of your qXR product? Could the product evolve into a holistic tool that detects every chest ailment, eventually?
qXR is a chest X-ray screening tool built with Deep Learning. It classifies chest X-rays as normal or abnormal, identifies the abnormal findings, and highlights them on the X-ray. qXR also generates a description of the X-ray findings, including name, size, and location of the abnormality, that is used to pre-fill radiology reports. The algorithms have been trained and tested using a growing database (over 2.5 Million) of X-rays from diverse sources.
As of today, qXR can detect up to 18 different findings with pre-read assistance that populates radiologist templates, has a worklist tool to segregate Normal and Abnormal Chest X-ray studies, and a proprietary algorithm that looks for signs of Tuberculosis on Chest X-ray. Our goal is to add to this list of 18 and eventually have qXR be the most comprehensive and accurate chest x-ray AL algorithm available to physicians around the world.
What is qER? How difficult was it for you to build a system from scratch that can diagnose a complex organ like the brain?
qER is designed for triage or diagnostic assistance in this setting. The most critical scans are prioritized on the radiology worklist so that they can be reviewed first. It detects critical abnormalities such as bleeds, fractures mass effect and midline shift, localizes them and quantifies their severity.
With this, our Deep Learning algorithms detect, localize and quantify a growing list of brain pathologies including intra-cerebral bleeds and their subtypes, Infarct, Mass effect, Midline shift, and Cranial fractures. Similar to our Chest X-ray algorithm, we used a large database of images and radiology reports to train and test this algorithm and we are getting great results. We are proud of the validation studies that we have done for qER including a peer-reviewed scientific publication in The Lancet, which is perhaps one of the most prestigious scientific journals in the healthcare field.
How do you see the raging trend of including ‘AI in everything’ impacting businesses?
While the changes that AI might bring are not yet fully realized, we know it has the capacity to change the business world, as we know it today. We are early on this journey and don’t fully understand how AI will ultimately impact all businesses. It’s logical to look at more routine AI-strategies that are already in use and can be mechanized as good kickoff points. This could mean anything from data analysis to manufacturing, essentially any business task that is repetitive. Already, across various sectors like healthcare and many others, AI-driven tools are driving better business decisions and outcomes.
The real value for society and for businesses would, however, come when AI companies are able to develop something that doesn’t exist yet. We are starting to see some signs of that. The majority of all AI in businesses today are focused on being able to complete tasks in a fraction of the time it would take several humans to complete.
Where do you see AI/Machine Learning and other smart technologies heading beyond 2025?
For healthcare in general, I think many of the concepts or early versions of technologies such as ingestible sensors, hand-held ultrasounds, wearable signal screens, etc. shall be real and in clinical use.
But, if we’re focusing on radiology, I love to indicate to an analogy to pathology to look at a test referred to as Complete Blood Count. In the 1980s, this was done manually, with pathologists having to identify cell patterns with the use of a microscope. It used to take 15-20 mins for this process. Today, machines can do 100s of those checks in mins, and pathologists can spend extra time on duties corresponding to detecting or grading most cancers. I imagine radiology could also be beheaded in an equivalent path. AI will assist the entire patient workflow as is defined today, perhaps take over some important yet non-value-added tasks like measurements, triage, etc. thereby permitting radiologists to concentrate on the vital and complicated situations.
What AI start-ups and labs are you keenly following?
I am a business guy and I like to follow companies that are using AI to solve a problem and not just AI labs. I obviously follow what Amazon keeps doing to integrate AI into every business process and so many startups trying to build self-driving cars using AI. The work coming out of Microsoft recently is also fascinating.
What’s your smartest work-related shortcut or productivity hack?
I have become a fan of tools like Slack and Calendly that have made my work more productive. As we work in a start-up, speed is of the essence in everything we do, from knowledge sharing to connecting with our customers. I use Slack not only for communication within the team but also as a reminder service to make sure I have someone looking into my To-Do lists apart from myself.
What would have been your alternate career choice if not into cutting-edge software sales?
I almost went into banking though chose early on that it was not for me. If you ask me today, my career choice would be to be a healthcare professional. I see healthcare as one of the most impactful paths one can take and sadly, one that has not been run professionally to get the best outcomes. I would like to be able to do that.
Thank you, Chiranjiv! That was fun and hope to see you back on AiThority soon.
Chiranjiv Singh is Chief Commercial Officer at Qure.ai where he is leading growth for this exciting startup and wants to bring the power of AI to healthcare, making healthcare more affordable and accessible for everyone. Chiranjiv truly believes that AI is the technology that has the potential to impact the Triple aims of global healthcare – quality, access and cost and is drawn to the challenge of making this potential come real.
Chiranjiv has over 20 years of experience leading commercial product and sales teams to success in diverse markets. His professional career has seen him work on many innovative product and business model launches – from the first US FDA approved cloud-hosted ECG interpretation algorithms and the first on-device AI algorithms for x-ray equipment. His last role was with GE Healthcare’s X-ray business where he led global marketing and commercial strategy. He has an MBA from IIM Calcutta and a B.Com (Hons.) from the University of Delhi.
From medical images to connected devices, the amount of medical data being generated is growing exponentially. This coupled with new forms of data such as the ones generated by genome sequencing and biosensors renders traditional diagnostic methods obsolete. At Qure, we are a team of passionate computer scientists, medical practitioners, and bioinformaticians who are on a mission to make healthcare more accessible and affordable using the power of deep learning. We believe that Artificial Intelligence (AI) is the key to ensuring that healthcare practitioners can focus on cases that truly matter, letting machines diagnose or treat the easier ones. Our aim is to leverage deep learning effectively to diagnose disease from radiology and pathology imaging and create personalized cancer treatment plans from psychopathology imaging and genome sequences. Our research is done in collaboration with several hospital chains, universities and research institutions. If you are a medical institution and would like to collaborate with us, please reach out to us.