The responses have been given by  Rohit Ghosh, Founding member, Qure.ai 

All around the world, people are facing unprecedented challenges and uncertainties as a result of the COVID-19 pandemic. As InnovatioCuris (IC), we are always on a lookout for healthcare innovations that are affordable and provide quality care. In the wake of this need of the hour, InnoHEALTH magazine scouted and interviewed some innovative startups to build an army of health transformers to mobilize and address this global health crisis.

Kritika Arora and Varsha Prasad interviewed Rohit Ghosh, Founding member,  Qure.ai on behalf of InnoHEALTH magazine.

By interviewing Qure.ai, we aim to understand and review the role of AI as a decisive technology to analyze, prepare us for prevention, and fight with COVID-19 (Coronavirus) and other such pandemics.

What are the reasons behind the company name , any story behind it?

Rohit, who is the founding member of Qure.ai, very honestly said “This is funny as there is no specific reason as such for naming the start-up as qure.ai. We were looking for a short and subtle name. It was basically that we were not able to get our domain name as cure.ai and that’s how we landed up with Qure.ai” !

What made you start the company ? Tell us briefly about your journey.

The primary idea initially was to apply the concept of machine learning and data science in the field of healthcare. Back then, deep learning was just a newly coined term, not hyped like now. So we were not looking at buzzwords but more on the real impact that we could create using technology.

Initially, we were exploring the different domains in healthcare to focus upon but very soon we decided to opt radiology as our field of interest. The primary reason was, we saw a problem that there is a huge gap in the availability of radiologists, globally. Moreover, I still believe the value that can be created in radiology is not only higher but more universal in terms of global protocols.

How is your company using AI? Tell us about the specialisations where it is being used.

As an AI for radiology company, we have launched two major flagship products in the last four year –

• qXR is AI for chest x- rays (qXR) 
• qER is AI for head CT scans 

 On the specialisation part, qXR enables the detection of  major abnormalities which are associated with chest x-rays out (of which TB is commonly used one in India and SE Asia and now COVID-19 as well). qER helps in automated reading of emergency findings from non-contrast head CTs. AI enables radiologists to analyse these X-rays and CT scans in comparatively shorter duration. It’s almost as good as radiology team having an extra albeit virtual member in their team working round the clock.

Explain us a typical day in office. How does an AI expert spend their day ?

Nowadays, it’s very tough to stay in office as the nature of work involves a lot of travel. But things were drastically different a few years back.

Very early on, he used to spend a good part of the day reading a lot of medical literature understanding basic medicine concepts and also research papers related to application of AI not only in healthcare and also in other domains.

 Later on things changed as he started devoting more time to developing solutions for the problem at hand. In addition to this , he also spends a lot of time on discussions with radiologists understanding different nuances. For example, understanding if they are getting false positives or negatives then what is the likely reason behind it and how it can be rectified. Working closely with clinicians was very helpful and is still of help, says Rohit. He also shared a personal learning by saying that he himself can actually read a head CT with a fair amount of accuracy!

Tell us the challenges you have faced or are currently facing in the development or implementation of AI.

There are a lot of challenges associated with development and implementation of AI in medical imaging. The first challenge is understanding medical concepts and grasping them in a relatively short time when you’re not from a medical background. Then the availability of data is the other big challenge for any AI organisation – in medical imaging it becomes even more pronounced. All of these are basically the primary challenges which they had to face in their first few years.

According to Rohit, the challenges which they faced after production of their AI products were different – one major challenge was how do they tackle the difference between the medical images across multiple sites. For example X-ray of chest would vary from demographics, machines etc. In addition to this , how these images are taken adds a lot of  variability to these images. 

Tackling all these challenges were critical to get to a successful working solution.

How can you overcome these challenges? Do you want to share any instances in the past where your team is able to overcome any particular challenge?

To this, very interestingly Rohit said “I would like to tell a very interesting story of myself only. I was working on detection of fractures via head CT scan. In the case of the skull fractures, it becomes very difficult to locate the fracture because there are several confounding impressions on CT and it’s extremely difficult to build an AI software for the same. What I did at that point of time was that I went to a radiologist to enhance my understanding about how he identifies fractures from a head CT scan and tried to mimic that in our AI tool. This helped in drastically improving the performance of software. So as I mentioned earlier, as a non-medicine folk trying to solve problems, the best way to overcome challenges is to work closely with an expert or an advisor. The solutions for other challenges were also unique in their own way.”

How reliable are these AI tools from a clinical perspective ?

The two things which determine the reliability of any AI based tool – the publicly available peer reviewed publications (as everyone can claim their own version of truth) and the testing of one’s own data (it works for others is not enough). So based on both the parameters one can easily get an idea of performance of any AI tool in a clinical perspective. Nothing more, nothing less.

What are the regulatory approvals your organization has?

Currently Qure.ai has CE approval for both the products. Rohit shared that they were the first organisation to have CE approval for screening 15+abnormalities from chest X-rays. The current CE certifications include all the abnormalities which can be detected by both the products. It was also exciting to know that they recently extended their qXR software capabilities to COVID -19 and are also having CE certification for the same .

Share the customer benefit with your solutions and the role AI plays in it.

The most important benefit because of AI is reducing the time in diagnosis of any kind of medical imaging. This can trickle-down in many ways to the end-user – sometimes this improves the productivity of the radiology team whereas other times it directly improves patient health.

Talking about COVID-19 since that’s the most interesting thing right now. The AI software helps in continuous and better monitoring of lung involvement for COVID-19. This helps in better decision making around treatment as well as discharge decision. In some places, the software is also used for screening of suspects (especially asymptomatic ones) in a mass screening set up. Mass screening setups can’t use RT-PCR as it’s scarce and expensive – that’s where Qure’s AI (qXR) along with Chest X-rays can help tremendously in saving lives.

In case of critical conditions like stroke or internal bleeding as well, the AI product for CT heads (qER) is a life-saver. As time plays a critical role in treatment of such patients, AI plays an instrumental role in reducing delays altogether.

Are there any general issues associated with AI products and services ?

Generalisability ,  interpretability and long-term accuracy are the main issues related to AI products.

What differentiates you from your competitors ?

On this, Rohit said “The major thing that differentiates Qure from our competitors is the quality of research which we carry out and thereby bring transparency to the overall process by publishing those results in peer reviewed journals consistently. This transparency makes us a favourite among clinicians who truly care for quality, helping us reach an incredible amount of global user base in just 4 years.”

Where do you see your organization in 10 years ?

The goal with which the start-up started with still stays and will stay for the next decade- to make healthcare more accessible and affordable for patients, more so in places with lack of infrastructure. 

Any brief message for our readers ?

There lies a huge scope for innovations in the field of healthcare, medical imaging is just a minor part of it. Counterintuitively, in low-middle income countries, the opportunities are much bigger as there are a lot of things that need to be fixed there as opposed to other places.

Here the catch is, that you just have to start looking for one problem that you think needs to be solved and want to work upon – from there multiple solutions emerge themselves. One last thing which comes from Rohit’s own story is that if you primarily don’t belong to the healthcare industry but are driven by the desire to impact, don’t be afraid. Give it time, be curious and it would work out. The only real skill is learning fast (whatever the topic be), rest everything falls in line.

With this, we wish all the luck and hope that in the coming decade, people see Qure as a path-breaking Indian AI company and that it creates a major global impact.

Interviewed by Kritika Arora and Varsha Prasad

The responses have been given by Udayan Dasgupta, Co-founder and Chief Data Scientist, Tricog Health

All around the world, people are facing unprecedented challenges and uncertainties as a result of the COVID-19 pandemic. As InnovatioCuris (IC), we are always on a lookout for healthcare innovations that are affordable and provide quality care. In the wake of this need of the hour, InnoHEALTH magazine scouted and interviewed some innovative startups to build an army of health transformers to mobilize and address this global health crisis.

Kritika Arora and Varsha Prasad interviewed Udayan Dasgupta, Co-founder, and Chief Data Scientist, Tricog Health on behalf of InnoHEALTH magazine.

By interviewing Tricog Health, we aim to understand and review the role of AI as a decisive technology to analyze, prepare us for prevention, and fight with COVID-19 (Coronavirus) and other such pandemics.

Reasons behind the company name, any story behind it?

The founders wanted to develop medical technology that could be used to diagnose complicated conditions,  like heart attacks,  early and even predict their onset.  They envisioned the product to be portable and easy to use and one that could be used by a  non-specialist to get accurate medical diagnosis anywhere, anytime. Being huge fans of StarTrek, they imagined that this modern-day medical TRIcoder  (https://en.wikipedia.org/wiki/Tricorder),  of  StarTrek fame,  could be designed by leveraging Artificial  Intelligence  (AI)  or  COGnitive science. By combining these two words (TRIcoder + COGnitive), Tricog was born – with the mission of magically delivering an expert medical diagnosis to the real world by leveraging the power of AI.

What made you start the company? Tell us briefly about your journey?

The founder,  Dr.  Charit  Bhograj, a  renowned interventional cardiologist,  routinely saw critical heart attack patients approach him hours after onset of symptoms.  And like other cardiologists,  he always knew that the clinical outcomes could have been significantly improved if the patients presented themselves earlier.  And this could have been possible if the patient had done a simple diagnostic test, called an  Electrocardiogram  (ECG).  This growing sense of frustration made him follow the patient’s journeys and realize that the root cause for the delay was often because the patient did not have access to a medical center that could conduct the ECG exam accurately, quickly, and affordably.

ECGs are the primary means of diagnosing serious cardiac conditions like heart attacks.  While  ECG equipment is fairly commonplace,  reading  ECGs requires the skills of a  cardiologist or an experienced physician.  Given that such expertise is limited in  India  (there are less than  3  cardiologists per million people in  India)  and in large portions of the developing world, misdiagnosis and delayed diagnosis are rampant. Furthermore, a delay of an hour in diagnosing heart attacks (golden hour) has been shown to significantly increase the risk of mortality and permanent damage to the heart. To make matters worse, cardiovascular disease is a  growing problem since it is linked to factors like age,  sedentary lifestyles, poor eating habits, and high levels of stress.

As machine embedded algorithms are only ~70% accurate in diagnosing heart attacks, most physicians prefer getting ECGs read by cardiologists, or, even worse, not keeping ECG machines as they cannot read it themselves,  further exacerbating the scarcity of getting a  quick and accurate diagnosis. This causes significant delays in diagnosing cardiovascular conditions and increases risks to the patient’s life.  The worsening state of the problem and the importance of the golden hour make the search for a scalable solution, to provide fast, accessible, and affordable cardiac care to the masses, especially important.

To solve this precise issue, he teamed up with a group of engineers, Dr. Zainul Charbiwala (a hardware geek), Dr. Udayan Dasgupta (an algorithms expert) and Abhinav Gujjar (a software guru) to start Tricog, a  company focussed on accelerating medical care and improving outcomes by enabling access to accurate cardiac diagnoses in an affordable and timely manner.

How is your company using AI? Tell us about the specialisations where it is being used and problems that you are solving?

The company believes in democratizing medicine by accurately diagnosing medical conditions in an affordable and timely manner.  In this journey,  they started off by introducing a  medical service focussed on accelerating cardiac care by providing instant ECG diagnosis for diagnostic ECGs captured from remotely placed ECG machines – ECG being the primary diagnostic test for detecting heart attacks and other critical cardiac conditions.

This solution consists of cloud-connected  ECG  machines placed at various remote centers,  which continuously push  ECGs to the company’s  Cloud where a  proprietary algorithm called THA first analyzes the ECGs and provides the preliminary interpretation for verification by the in-house team of cardiac specialists who are present   24/7/365  company’s centralized   ECG   Analysis   Hub.   The physician-verified reports are sent back to the remote center via  SMS/APP,  who can then coordinate with neighborhood hospitals to provide further care.

They are currently in the process of using a similar approach to develop solutions for providing instant diagnosis for treadmill tests  (TMT)  and echocardiography tests.

Explain us a typical day in office, how does an AI expert spend their day? Tell us some under the hood details?

Developing algorithms to solve challenging real-world problems is an extremely enjoyable journey.  It typically starts off by clearly defining the problem,  in terms of possible inputs and their expected outputs, followed by extracting features from the data, annotating the data, and visualizing it in multiple ways to gain insights. This stage also involves working with domain experts (in this case, cardiologists) to understand how they diagnose  ECGs.  Depending on the amount of data available and armed with the insights gained from the previous data exploration stage, the AI engineer then starts evaluating various approaches to solve the problem  –  be it signal processing methods,  machine learning approaches, or deep learning models.  The data is typically split into training and test data sets so that the algorithms can be trained on the training data and their performance evaluated on the unseen test data sets. This stage can also involve reading and implementing published papers that have solved similar problems. The work is typically iterative in  nature as approaches are tried,  evaluated, and modified to improve performance on the test data sets.  Once the engineer is satisfied with the performance,  he may do further optimizations and architecture changes for efficient real-time deployment and a final round of validation with domain experts on large unseen data sets. When the domain experts and AI engineers are satisfied and the solutions are deployed,  the  AI  engineer picks up the next unsolved problem and restarts the journey.  Overall depending on the stage,  every day in this journey is different as the engineer struggles with a  series of puzzles and challenges along the path  –  but the sense of joy and achievement when each of them is solved is well worth the effort, says Mr. Dasgupta.

Tell us the challenges you have faced and are facing in development/implementation of AI?

Implementing effective  AI  in the area of healthcare,  which can provide real value to physicians,  has several challenges. A few of them are as follows:

• The promise of AI stems from the fact that it should improve with more data, but collecting such vast amounts of real-world data can be very difficult. People typically use a service (and share data)   with an expectation of a  result and especially in the initial phases when the  AI  is rudimentary,  providing them with accurate results which would allow the service to scale (and lead them to have the confidence to share more data), becomes very difficult.
• Several medical diagnoses,  including reading  ECGs,  is a  mixture of art and science and can be subjective. Compounding this issue is the fact that ground truth can be hard to get and is often substituted with physician’s opinions that can vary across physicians.  This can lead to large interobserver variability and getting universal reference data for training algorithms can be difficult.
• Cardiac abnormalities occurring in the broad population have a lot of variations (there are over 200  abnormalities that can be picked up by an  ECG),  and the data distribution is extremely skewed.  This means that several conditions are very rare and the data collection process for such conditions is typically slower, making data-hungry AI algorithms struggle. And of course, the overriding concern and challenge are that given the nature of the task, the AI has a very small margin of error given the large implication of incorrect decisions.

How can you overcome some of these challenges? Do you want to share any instances where in the past your team was able to overcome any challenges?

The first challenge was devising a model that would continuously improve in performance over time. By using the hybrid model  (Human  +  Machine  Intelligence)  they could roll out a service through which large amounts of data could be collected,  and the algorithms improved.  The human cognitive load reduced as the machine algorithms got better. According to Mr. Dasgupta, all throughout the process the end-customers always got accurate human-verified diagnoses, allowing people to believe in the diagnosis and the service to scale.

The second challenge, namely interobserver variability, was a very tricky challenge. Their approach to solve this issue was threefold:

• Standardized reading procedures and rules followed by doctors,
• Developing a proprietary algorithm approximating these rules which also provided evidence to the doctors to explain the algorithms’  conclusions.  The viewing system that incorporated the algorithm results were carefully designed to promote usability while combating automation bias to keep the Human-Machine model accurate.
• Implementing rigorous quality control protocols coupled with continuous testing and training to help physicians adhere to the standardized reading procedures.

Implementing AI algorithms that could work well with smaller amounts of data and provided improved performance when larger amounts of data became available, involved designing various algorithms to solve the same problem, with varying dependency on data size. The first version of algorithms typically used little data and performance was improved by hand-tuning.  As more data was available,  other data-hungry algorithms, whose performance naturally improved with more data, were brought into the fold and combined with the earlier set of algorithms to improve the overall performance.

In summary,  by building out this technology backed medical service, the start-up developed and deployed a business model which could deliver accurate results from the very beginning, but one whose efficiencies would continuously improve with more data. It was always apparent that a human-only approach would not scale and would be prohibitively expensive.   By using the hybrid model  (Human  +  Machine Intelligence) they rolled out a service through which large amounts of data could be collected, and the algorithms could be improved.  Through this process the human cognitive load reduced and system efficiencies improved over time while maintaining high overall accuracy of the hybrid system at all times.

How reliable are these AI tools from a clinical perspective, tell us about the regulatory approvals, which stage are you in currently?

The start-up’s solution is currently deployed in over 3000 centers in both rural and urban locations and has diagnosed  over  3.5  Million  patients  with  50% being abnormal and over 120000 of them being critical. They have recently implemented their first statewide STEMI program in Goa, and are working with multiple other state governments on similar programs. Outside of India, they also have rapidly scaled services in South East Asian countries (Philippines, Malaysia, Indonesia) and Africa (Kenya, Nigeria, Cameroon), shared Mr. Dasgupta.

The efficacy of the human-AI based platform has been key in creating this outsized impact. They are currently pursuing regulatory approvals for their solutions so that they can market these tools in other markets.

It was interesting to know that the algorithms have surpassed the performance of FDA approved algorithms in head-to-head testing on real-world resting ECGs containing over 200 abnormalities. Their reliability from a clinical perspective is evaluated continuously, out of  ~150,000 ECGs they diagnose every month are verified by doctors. In such real-world tests it has been seen that doctors only modify about  1  out of every 5 diagnoses that the algorithm produces. The AI is expected to get even better as more data is collected, especially for rarer conditions.   These efficiency gains are a testament to the fact that over the last 24 months, the ECG load has increased by ~10x while limiting the medical team growth to less than 50%.

Share the customer benefit with your solution, and the role AI will play?

Doctors, the customers, value the company’s services because of its accuracy and quick turnaround times, says Mr. Dasgupta. AI has been pivotal in meeting this expectation while keeping the service affordable.  With the use of medical sensors becoming widespread,   large amounts of data will be collected and it will be practically impossible for human-only approaches to analyze the data, derive insights, and develop treatment plans while keeping the process timely and affordable. AI will be increasingly adopted to help with these tasks. It will start with the simpler tasks of data analysis and will help physicians derive insights and then plan treatment.  Over time,  based on human interactions,  AI  will start taking on more complicated tasks, thereby increasing efficiencies of the healthcare systems.  However in medicine,  given the variety of complexities of the human body and the subjectivity of the diagnosis,  human judgment will remain important – making physicians and AI work in tandem, especially for complicated cases.

Are there any general issues associated with AI products and services?

The main issue in the adoption of AI in healthcare is the concerns (often valid) about trusting technology for taking critical decisions, especially given that the impact of an incorrect decision is large. This lack of trust is not helped by the fact that some of the recent AI techniques, like deep learning, are typically opaque,  i.e. although they may deliver the correct diagnosis, they are unable to explain the results for the physician to understand its reasoning. And although some methods work very well when the unseen data is similar to past data, they may struggle to generalize learnings over data from disparate sources.

What efforts are you making to overcome associated disadvantages?

The start-up has never used AI to independently diagnose critical heart conditions and has rather used it as an assist tool for their in-house physicians. Also, throughout this journey, they have focussed on two aspects:

• Always providing visual evidence to the physicians which justifies the algorithms’ decisions
• Always learning from the physicians’ reviews, especially when they differ from the algorithm’s conclusions so that over time the algorithms improve and better mimic human decision making. With these guiding principles and constantly working with the physicians, the end-users, the algorithm has been able to gain their trust for the majority of diagnosis and also their patience as they work with the algorithm designers on the remainder of the conditions where the algorithm continues to improve.

What differentiates you from your competitors?

The virtuous Human <=> AI cycle embedded in Tricog’s business model has been key in building a robust algorithm platform tested on real-world data which has provided increasing value to the business over time.  Their algorithms started off by leveraging a  few public domain databases and using signal processing based approaches for feature detection and measurements, coupled with an expert system built in consultation with the in house medical team.  However as their  ECG  diagnosis service became popular,  it helped in collecting large volumes of high fidelity real-world data which were annotated in real-time by the in-house medical team. 

Another key advantage is that the solution was always architected to be cloud-first, which is in sharp contrast to even high-end traditional ECG machines with in-built diagnosis algorithms leveraging sophisticated, but fixed, signal processing techniques. By moving the data and computations to the cloud, a few interesting benefits occurred,

• The computational resource constraint was removed
• The algorithms could remain dynamic and change over time
• The aggregation of data allowed machine learning and deep learning approaches to be incorporated within the product.
• The elaborate ECG viewing system allowed a lot more supplemental information to be shown to the cardiac specialists, leading to a quicker and more accurate diagnosis.

Being the first physician verified instant ECG diagnosis service, they seem to have effectively identified and effectively solved a deep problem in the healthcare landscape.

Where do you see your organisation in 10 years? Any brief message for our readers?

The start-up intends to provide technology accelerated solutions to real-world medical problems. It has started off by making cardiac care accessible, affordable, and timely. Developing solutions for the Indian market has successfully led them to take its solutions to countries in Southeast Asia and Africa. Going forward they intend to grow these services in other geographies while developing differentiated services for delivering quality healthcare at home.

Interviewed by Kritika Arora and Varsha Prasad

The responses have been given by Dr. Manjiri Bakre, CEO & Founder, Oncostem Diagnostics

All around the world, people are facing unprecedented challenges and uncertainties as a result of the COVID-19 pandemic. As InnovatioCuris (IC), we are always on a lookout for healthcare innovations that are affordable and provide quality care. In the wake of this need of the hour, InnoHEALTH magazine scouted and interviewed some innovative startups to build an army of health transformers to mobilize and address this global health crisis.

 Kritika Arora and Varsha Prasad interviewed Dr Manjiri Bakre, CEO & Founder, Oncostem Diagnostics on behalf of InnoHEALTH magazine.

 By interviewing Oncostem Diagnostics, we aim to understand and review the role of AI as a decisive technology to analyze, prepare us for prevention, and fight with COVID-19 (Coronavirus) and other such pandemics.

Reasons behind the company name, any story behind it?

The company focuses on Oncology and was built to develop tests to predict cancer relapse, to help personalized cancer treatment planning. Taking “Onco”, from oncology and “Stem” inspired by cancer stem cells which are known to play a role in cancer relapse. Hence the name OncoStem.

What made you start the company? Tell us briefly about your journey?

OncoStem was founded in 2011, after being moved by the death of a close friend due to breast cancer. She died despite being diagnosed with early-stage disease, maybe due to a limited understanding of the tumour biology driving the disease. Companies in the West were developing tests that could analyze tumour biology and determine the risk of relapse in patients with early-stage breast cancer patients. These tests could help personalize a patient’s treatment plan based on that patient’s individual risk of relapse. Since these tests are very expensive, Dr Manjiri shared that she approached doctors in India and pitched her idea of developing a home-grown and affordable version of such a test. And that’s how the journey started.

How is your company using AI? Tell us about the specializations where it is being used and problems that you are solving?

OncoStem’s flagship product “CanAssist Breast” is a machine learning-based prognostic test that helps personalize treatment for early-stage breast cancer patients who are hormone receptor-positive (HR+) and HER2-negative.  The standard treatment for HR+ HER2-negative breast cancer involves both chemotherapy and hormone therapy. The risk of relapse in early-stage (Stage I and II) HR+ HER2-negative breast cancer is very low (10-15%) even if patients are given hormone therapy alone. This implies a majority of patients (~85%) are being overtreated with chemotherapy, which has toxic side effects and lowers the quality of life of patients.  

The company uses cutting-edge machine learning-based algorithms that predict the risk of recurrence for every patient. Machine learning is known to be a more advanced tool to develop prognostic tests where patterns of patient information need to be understood and analyzed.  It maximizes diagnostic accuracy thereby improving patient outcomes. Machine learning-based methods also have flexible “transfer functions” which allow them to model complex processes such as tumor recurrence.

Explain us a typical day in office, how does an AI expert spend their day? Tell us some under the hood details?

At work typically, Dr Manjiri wear multiple hats in one day depending on a day, it can be a sales meeting first thing in the am, followed by scientific publication review, followed by issue in purchasing a reagent, to trouble shooting a Statistical analysis work or visit a hospital to pitch the product for commercial purposes or propose a study with a hospital, help design a brochure for a conference, meet investors, interview candidates to reading and discussing a legal document with our lawyers

Tell us the challenges you have faced and are facing in development/implementation of AI?

Conducting validation studies in India can be challenging because the clinical trial framework is not as well established as developed countries. Dr Manjiri shared that they must have approached 50-60 hospitals to work on developing and validating CanAssist Breast, which ultimately led to 10 hospitals signing up. There are many reasons for this – 1) one is that the doctor-patient ratio in India is well below the developed world. Each oncologist in India sees many more patients per day leaving them with little time to focus on research studies. 2) Lack of documented patient records and follow-up also makes it difficult to conduct studies that require patient history. Patients often move back to their hometowns after treatment and are lost to follow-up. Availability of data is crucial to AI/ML. A machine learning algorithm can only be as good as the input data. Robust data makes for a robust algorithm, otherwise its garbage in, garbage out 3) Lack of enforcement and encouragement of importance of research studies, clinical trials, detailed documentation from government and regulatory authorities. If they enforce and encourage these activities,says Dr Manjiri, they can develop far more world class diagnostics, drugs and medical devices than they are currently doing.

How can you overcome some of these challenges? Do you want to share any instances where in the past your team was able to overcome any challenges?

Through persistence and perseverance, Dr Manjiri found clinicians who believed in her cause. They were passionate about working on new technologies, who devoted time and effort to participate in our validation studies. Without them, this would not have been possible.Her early investors, Artiman Ventures, were also very helpful in making connections to clinicians in India and the US.

How reliable are these AI tools from a clinical perspective, tell us about the regulatory approvals, which stage are you in currently?

Specifically, in healthcare, ML has led to exciting new developments that could redefine cancer diagnosis and treatment in the years to come. ML can increase access to treatment in developing countries which don’t have enough cancer specialist doctors to deal with the rising incidence of cancer that can treat certain diseases, it can improve the sensitivity of detection, add more value in treatment decisions of cancer, and it can help personalize treatment so that each patient gets the treatment that’s best for them.  In many cases, they can even add to workflow efficiency in hospitals. The possibilities are endless!

To share some examples, LYNA (LYmph Node Assistant) by Google detects spread of breast cancer metastasis early and can reduce the burden on Pathologists as well.  A deep learning convolutional neural network or CNN – developed by a team from Germany, France, and the US can diagnose skin cancer more accurately than dermatologists. In a recently reported study, the software was able to accurately detect cancer in 95% of images of cancerous moles and benign spots, whereas a team of 58 dermatologists was accurate 87% of the time. 

The move from lab to actual practice has happened already for some AI-based solutions such as the FDA-approved imaging tool called IDx-DR for diagnosing diabetic eye disease.

The test CanAssist Breast is CE marked and is performed in a lab which is ISO 13485 certified, NABL, and CAP-accredited.

Share the customer benefit with your solution, and the role AI will play?

The start-up provides oncologists with a well-validated and affordable solution that can be used to improve the quality of life of patients.Their machine learning-based algorithm calculates a risk score that is patient-specific. This score decides if patients can avoid chemotherapy. Chemotherapy has various side-effects, both short term and long term. In cases where its benefit does not outweigh the risks, it can lead to a serious decline in quality of life. The current treatment guidelines require oncologists to give patients optimal treatment based on their individual risk profile. But the tests developed by Western companies are very expensive and out of reach of most Indians. CanAssist Breast gives oncologists an evidence-based and affordable way to decide which patients can be spared the burden of chemotherapy.

CanAssist Breast is based on machine learning that is very well suited to building survival models in cancer because it improves patient outcomes by maximizing accuracy.

Are there any general issues associated with AI products and services?

Dr Manjiri said, technologies like AI indeed have tremendous potential and all stakeholders like the promising algorithms, accurate clinical and relevant in vivo data, clinicians, institutions have to align themselves to reap meaningful benefits from it. One must remember that excellent technical innovations in AI can not fix social/political problems. Also, the data input to AI must be in high volume and of clinically high quality/relevance. Fundamentally flawed data can not substitute for high volume. Currently, most of the AI applications are using the paradigm of ‘deductive reasoning’ and they need to move from there towards ‘inductive reasoning’.  

She also shared that they have traveled a fair amount in the AI path to excellence but one must be cautious going further to embrace the brilliant promise it holds. What they need next is to move from theoretic benefit and evangelical sales to established use cases and robust, clinically-relevant data.  

What efforts are you making to overcome associated disadvantages ?

The startup has established the accuracy and clinical utility of their test through years of rigorous validation. This has helped prove that their technology works and adds value in the clinician’s workflow. They are also working on additional studies, both locally and internationally, where long term data is available.

What differentiates you from your competitors?

According to the CEO and founder of the startup, their product, CanAssist Breast is the only test to be validated on Indian patients. Competitors have developed and validated in Western countries, on Caucasian patients. Caucasian breast cancer patients tend to be older and post-menopausal at diagnosis, whereas their counterparts in India tend to be younger and premenopausal. They also have other USPs:

1. Only product to have data on Indian patients. 
2. Is affordable as compared to expensive Western tests.
3. The turn around time or TAT is claimed to be the shortest so they report within 8-10 days compared to 3 weeks taken by some of the competing tests. 

Where do you see your organisation in 10 years? Any brief message for our readers?

OncoStem is working to globally validate its “Made-in-India” test CanAssist Breast, by undertaking studies across multiple continents. Also working to build a pipeline of tests that can personalize treatment for other cancers like ovarian cancer.

Working with pharmaceutical companies to develop targeted drugs based on the learnings from CanAssist Breast, which can help treat those patients who don’t respond to conventional therapy. Drug development is a long process and they hope in the next 10 years they see major activity on this front.

Interviewed by Kritika Arora and Varsha Prasad

The responses have been given by Suthirth Vaidya, Co-Founder at Predible Health

All around the world, people are facing unprecedented challenges and uncertainties as a result of the COVID-19 pandemic. As InnovatioCuris (IC), we are always on a lookout for healthcare innovations that are affordable and provide quality care. In the wake of this need of the hour, InnoHEALTH magazine scouted and interviewed some innovative startups to build an army of health transformers to mobilize and address this global health crisis.

Kritika Arora and Varsha Prasad interviewed Suthirth Vaidya, Co-Founder at Predible Health on behalf of InnoHEALTH magazine.

By interviewing Oncostem Diagnostics, we aim to understand and review the role of AI as a decisive technology to analyze, prepare us for prevention, and fight with COVID-19 (Coronavirus) and other such pandemics.

Reasons behind the company name, any story behind it?

Predible comes from predictions that are credible. It is basically a shortened form of credible predictions through the use of Artificial Intelligence. Healthcare is all about credibility, credibility through which a doctor is looked up as a doctor. So, fundamentally to core guidance of principles about what predictions should be, one key adjective should be credible to describe it. Predible health is essentially credible-predictions.

What made you start the company? Tell us briefly about your journey?

Mr Suthirth Vaidya shared with us that he along with Mr Abhijith Chundru started the company right from IIT Madras in 2016. In 2015, both worked on research projects that they built without deep learning for neuroimaging, and this was basically from 2014-15. Very early on, they discovered methods to use deep learning for 3D imaging. He also shared that they won the Research Challenge Award organized by John Hopkins University in New York. This was for the detection of a particular condition called Multiple Sclerosis on brain Magnetic Resonance Imaging (MRI) and was also covered in the media. From that point,  2014-15, they realized they have this great technology that works for 3D MRI images, but there are so many more problems in the radiology field where this technology is applicable.  Given that large amount of data is the crucial ingredient for really powerful algorithms,they figured India would be at an advantage. Basically, the point that they figured out the technology at the right time, as well as the fact that there is a lot of data in India, lead them to start the company.

How is your company using AI? Tell us about the specialisations where it is being used and problems that you are solving?

The application of AI in the field of radiology is mainly the core competence and hence dealing primarily with the analysis of images like X-Rays, CT-scans, and MRI. The current flagship product called LungIQ is basically the analysis of  Lung CT images like 3D lung images. The Algorithms can detect multiple conditions, such as early cancer detection, COPD quantification, and detection of interstitial lung diseases. And the way the organization approaches products is really helping pulmonologists deliver better care and how they need to add in the workflow so that radiologists can deliver the analysis in their everyday life. Therefore, not only that AI is offering the knowledge that can benefit patients, but that we can all communicate better and use this information in everyday practice, both for radiologists and the specialist.

Explain us a typical day in office, how does an AI expert spend their day? Tell us some under the hood details?

As you already know, building and developing good AI and designing AI Algorithms is 20% of the work and about 80% of the work is on Data and Data Curation. And on a typical day, AI specialist job is going to be more like the data scientist looking at data and finding a few different types of cases like which algorithms don’t perform well and trying to find some of those data, getting it back into training and trying to move the algorithm to higher and higher accuracy levels. Right now, the organisation has COVID-19 Algorithms and every day they see more and more different types of cases of COVID-19 patients that they are trying to figure out and rigorously test and see if the algorithms are robust to all kinds of patients we see. Mr Vaidya shared that as soon as they see the case where they struggle, they seek to go to find more similar cases, and write to the radiologists with whom they work, and understand the data and problems. It’s always about finding gaps and sampling data better so that the algorithm works better.

Tell us the challenges you have faced and are facing in development/implementation of AI?

The difficulty in implementing AI is that it’s such a new area, and the moment you go to a hospital or a doctor and use those two terms, the kind of response you get is right from enthusiasm to interest to scepticism. “And, basically, we ‘re trying not to think too much about the fact that we’re using AI, we ‘re not going to make it appear that we’re using AI under the hood in any discussion until it’s practically asked. The picture of the product is the value that we are offering. Much of our conversations with doctors are about what the product really provides in terms of quantification, precision, and how it fits with the workflow and things like that” says Mr. Suthirth. Integrating these things into enterprise-wide systems was initially a kind of challenge, ensuring that the tools built to fit into the clinical workflow without having to jump to too many screens. They have had the initial challenge of providing access to a vast volume of data from hospitals addressing data privacy concerns, data protection issues and, in particular, India does not have very specific rules or legislation on data privacy and data sharing. Mr. Suthirth also said that they ensure they have committee approvals and institutional board approvals, and all of this before they can have access to massive hospital data.

How can you overcome some of these challenges? Do you want to share any instances where in the past your team was able to overcome any challenges?

In terms of challenges, where the company started 4 years ago, so many open source databases are readily available today. If you want to build a chest X-Ray Algorithm or an instance now you have literally hundreds of thousands to three hundred thousand chest X-Rays available to start training more instantly. So, the obstacles to starting something and joining the market have finally gone down. But, if you want to get the data from big hospitals, you also have to go through the process, so you need a clinical champion to work with you and get professional approvals from the management team. Yeah, sadly, there’s no easy way out other than being able to use existing open-source data to get started, but what’s more, it is still all about the customer values that you offer and ensuring from day one that you have seamless integration into the workflow. It’s simply a matter of talking to hospitals and doctors before you go out and develop the product, says Mr. Suthirth.

How reliable are these AI tools from a clinical perspective, tell us about the regulatory approvals, which stage are you in currently?

From a clinical point of view, it depends on how you actually position the product. For example, any of the reports bypassing the signature of a radiologist is never recommended. So, the radiologist has to control all the results that are generated and sign on the reports before approval or they go out in the hands of patients or specialists. It is therefore very important to keep radiologists in control of all reports and not to bypass how the product is delivered to the hospital. It’s a critical point on how to design and implement validation tests for these Algorithms as well because what you really have to aim for is that AI really increases the doctor and makes him quicker and doesn’t require him to spend more time or lead to unnecessary biases. To that, the organization has done multi-reader studies where they compare the studies and give them to radiologists to read with AI assistance. He also shared that they are working with the regulatory bodies to obtain a CE-mark for the products.

Share the customer benefit with your solution, and the role AI will play?

Taking the current example; the start-up is working in the COVID-19 scenario, and what they are doing in this COVID-19 is if you look at the CT scan and send it to them, they ‘re going to measure the severity of the disease in the lungs. So now, if you do the RT-PCR check, you’ve been told if you have COVID-19 (yes or no) but you don’t really know how much your lung is compromised and what extent and level you are until you do the radiology clinical examination. However, if you feed it to predible’s solution, it will read the lungs and give you a percentage of how much your lung is affected. It will help the doctor track more overnight how the patient’s condition progresses and whether it deteriorates and determine whether or not he needs to be admitted to I.C.U. or can be discharged. So, it offers a percentage, a more objective way to track the success of the study. Examples can be replaced by many chronic respiratory conditions such as idiopathic pulmonary fibrosis, COPD, the solution can study lung overtime patterns such as how these chronic patients perform, quantify how their lung functions over these windows, and how much better they get.

Are there any general issues associated with AI products and services?

To this, he said, as such there are no general problems. Everyone’s doing AI these days, and there are a lot of positive and better solutions. Some hospitals and clinics may not know how to pick the best solutions. Having a lot of players helps the industry move forward, and the people who are trained, know what to look for. Apart from market education, clinicians and doctors need to understand how AI works and how AI needs to get into the workflowOf course it will take some time for early doctors to use it and speak about it after sharing their experiences, sharing it with communities, and slowly trickling to more sites. That is the process of market acceptance, so all just have to wait for it as time passes.

What differentiates you from your competitors?

Working in the lung space and pulmonary space; actual competition is not really the other radiology AI provider. Presently, we are not seeking the fight/competition with others. The market adoption of AI, in general, itself is so minuscule today that it is too early to be in a competition, says Mr. Suthirth.

Where do you see your organisation in 10 years? Any brief message for our readers?

Saying “Ten years is a long time, and the clinical operating system is where we want to go” Mr. Suthirth continued that they ‘re looking at one specialty lung pulmonology, and they’re looking to help them better assimilate information on radiology by using AI over time, which they want to expand into many specialties, and want to be able to use many more data sources beyond radiology. He also said that “We want to build the suite of digital products that could be enterprise and mobile-first, that doctors can consume tomorrow. We want to create medical products that make up the hospital’s footprint, help them simplify the clinical process, provide quality treatment – both evidence-based and personalized patient care. The data elements are how we are able to use all the data and how we are able to gather evidence-based data to give each person better treatment. This is the perfect structure that we want to create.”

We hope our readers get a great insight about the current trends and future of AI and how Predible Health is working during this COVID-19 crisis. 

Interviewed by Kritika Arora and Varsha Prasad