Startup Saturday: AI,IoT, ML; high-end solutions put to work
These biz heads stay ahead of the curve, harnessing tomorrow’s upgrades today
Niramai Health Analytix
Tech-ing mammograms for early diagnosis
Geetha Manjunath, founder, computer scientist with a PhD from IISc; spent 25 years leading research teams at Hewlett Packard and Xerox ; company operational from January 2017, incorporated in July 2016
Funds
$7 million, in two rounds.
Both VCs and angels
Lead investor for seed round: Pi ventures
Lead investor for Series A: Dream Incubator
In the beginning
Manjunath’s cousin was diagnosed with breast cancer in 2014, she died in 2016.
“That was the moment I felt I should do something about this disease. The current mammogram tests do not work on younger women and I thought, how could we solve that issue?
“In the past, thermal imaging diagnosed breast cancer, as an adjunct modality, as these thermal images are very complex and error-prone when analysed visually.
“We decided to use this modality along with data science to develop a test that would be no touch, no radiation; and more reliable.”
How it is done
“In mid-2014, we started to work on this problem. Our solution works with thermal camera of multiple vendors. Currently we import it from Sweden and Korea
“Thermal imaging measures temperature distribution on the skin using an infra-red camera. “Abnormal tissue activity can be detected as it gives out higher temperature than normal tissues.
“We measure around 4,00,000 temperature points on the chest. What we did was to combine data science with thermal imaging to create a test that detects breast abnormalities early and we call that Thermalytix for Thermal Data Analytics.
“We analysed this data using ML and computer vision techniques. Using historic information as training data we built the ML model. This training data included patients diagnosed with cancer (malignant); those with an abnormality, but not cancer (benign); and those normal.
“During training, the ground truth for our models came from standard tests - mammography or sono-mammography and then biopsy, which helped train our machine learning models accurately.
“In addition to thermal images, they also had to take into account various other patient features such as family cancer history, demographic info, patient complaints, and so on.
“We even hand-crafted special thermal abnormality features, that are patented now.
“There are 117 abnormal patterns and we made a probability model of these features so that for a new person it could predict the likelihood of malignancy
“Since AI is built on past data to predict the future, Manjunath had to tie up with hospitals to exclusively fine tune the models in a clinical setting, where the treatment itself is not based on the output of this machine.
Niramai has 1.25 lakh thermal images in its data set.
To market, to market...
In 2018, trials showed promising results that Thermalytix - thermal imaging enhanced with AI – worked better than a typical mammogram for younger women.
“Our diagnostic tool was a no contact. Non-invasive. Non radiation. Women simply had to sit in front of the machine and in 20 minutes you got to know whether you have a malignancy or not.
In fact, our results so far show that the test is 20% more sensitive than a mammogram.
A mammogram cannot test young women because their breasts are dense. Our imaging can. And because it does not use radiation, it can be repeated many times without any side effects.
To patent, to patent...
Like any innovation developed by a company Manjunath got patents in place.
“Doctors are generally reserved about new technologies. They don’t want to take any risk with something new. We had to do a lot of evaluations for doctors. They would want us to show how it worked on a patient with say an abscess or at different stages of cancer, which we did, before they accepted it.”
RoI
The return on investment (RoI) comes from hospitals and diagnostic centres.
“Today we have tested 27,000 women in multiple locations and have our tests available in 50 centres.”
Since the machine is portable it has been offered to rural areas - where a woman can get notified if she has a malignancy or an abnormality that needs her to go to a tertiary care centre.
“The focus now is to make our tests available to everyone. Because we use cloud computing, the test itself can be conducted by less skilled workers and reviewed and certified by experts remotely - that helps in scaling. It is affordable, quick, painless and radiation free.
VeCrear Technologies
Veena Samartha, 25 years technology experience
In 2018, Veena Samartha and Ganesh KJ co-founded VeCrear Technologies with the aim to provide IoT solutions to four industry verticals – consumer, telecom, automotive and manufacturing.
Funding
Self-investment of Rs 1 crore
Key numbers
Turnover of $1million
43 full-time employees
In the beginning
“We had a company earlier where we worked on mobile solutions. We worked extensively with enterprises in the FMCG sector. Since we had the experience of software, hardware and wireless development, as sensor technology became more accessible, sharpening our focus onto internet of things (IoT) was the best way to utilise our strengths.”
“There are many use cases and scenarios like location-based services, that help track and monitor mobile assets and personnel; provide various sensors, alerts and real-time preventive actions ensuring home safety; and automating business operations for consumer service providers. “
“Our technology can help enterprises by providing customised solutions and real-time data analytics.”
How it works
“What we did initially was to get experts as consultants. With our experience and contacts in the industry we were able to tap such talented & experienced people. In cases where niche skills like hardware fabrication and manufacturing was required, we had tie-ups with companies that have relevant expertise.
“With technology experts and eco-system partners in place, VeCrear was ready to dive into the market.”
A case in point is a food chain that VeCrear serviced.
“We had suggested a human machine interface (HMI) with a connectivity module to the customer to interface the vending machine with a connected device. e built an initial HMI-based proposal for the customer, the customer was not happy with the overall solution cost.
“We thought why not use a tablet as the user interface? This required further tweaks to ensure charging and data transfer could happen through the same USB port with a tablet enclosed in a heat and moisture proof casing. This solution not only solved the original problem of monitoring, but also provided additional benefits of personalisation, ad streaming, supply and maintenance monitoring, by having a centralised cloud back-end. The tablet considerably reduced the interface cost.”
To market, to market
Competing with large corporates who have money and talent at their disposal meant having a clear strategy.
Large companies can easily move people from various departments and skill sets to work on a project.
However, being small meant being nimble and quick. We could customise or modify the original offering much faster than the large companies. If a customer wanted a specific solution, we could bring it to life much faster and with higher efficacy. We work with our customers right from understanding their problem statement, designing and through development.
Tech change, take change
“One can respond to the fast changing technologies only if you have good understanding of the depth and breadth of technology yourself. As a tech entrepreneur one should be able to look at the bigger picture and not just at the narrow view of the area in which you operate. Once you get that perspective it is possible to move at the speed of new tech. On the same lines, we are currently focusing to enhancing our IoT platform with AI/ML assets.”