Pathology Outliers

SigTuple observed poorly staffed and overworked labs in India, and knew that it could speed things up cheaply

RA Chandroo

Like many other start-ups, the story of SigTuple started in the drawing room of one of its co-founders, in July 2015. But unlike many of its peers, the three-year-old start-up has marquee investors, such as Accel Partners, Chiratae Ventures (earlier IDG), Bansals of Flipkart and Axilor, founded by Infosys founders, on board. While many of them came in during its Series A, they stayed on and ponied up the entire amount for Series B. This included even the Bansals, who had come in as angel investors. 

They took lesser time deciding on the Series B round for $19 million — raised this June and led by Accel Partners and IDG Ventures — compared to Series A for $5.8 million, in February 2017. It took six months to close the first and three months for the second.So, why are its investors so excited about SigTuple?

Ranjith Menon Executive director, Chiratae Ventures

“What gets us excited is the team’s fresh perspective on the age-old diagnostics space. They are deploying new-age technologies such as AI, while not making significant changes to existing clinical workflow. Therefore, adoption is easier,” says Ranjith Menon, executive director, Chiratae Ventures. 

Manish Singhal, founding partner, Pi Ventures adds: “They have digitised a process that was being done through chemical processes all these years. They have built an AI engine and a platform that can do haematological tests that are non-trivial.” Founded by former American Express executives Rohit Kumar Pandey, Tathagato Rai Dastidar and Apurv Anand, SigTuple digitises the testing process for labs, hospitals and clinics. 

SigTuple’s digital scanner, named AI 100, digitises any kind of biological sample including blood, urine and semen. This data is then pushed to a cloud-based platform, named Manthana, on which AI algorithms run. The algos analyse the digital data and generate a report, which can be reviewed by a pathologist located even in a remote location.

“This means, if I deploy this device in a tier III city, one can just put the slide in the device and click a button that says ‘scan.’ The processing and report generation will be done on the cloud and a technician or a doctor sitting in a tier I city can then do the diagnosis. No shipping of sample necessary,” explains Pandey. A report is generated in less than 10 minutes.

This report is visually friendly too. All the numbers are converted to representative images, for easy understanding. “Say ‘x’ number of white blood cells (WBCs) have been found in your blood sample, we show the images of those ‘x’ number of WBCs,” he says. That means a lot of pictures, which could slow down the download, but the SigTuple team has fixed that too. “Algorithms optimise the size of images,” he adds.

Ramesh Byrapaneni, Managing director, Endiya Partners

Ramesh Byrapaneni, managing director, Endiya Partners, who has been a cardiologist for nearly three decades, sees great value in SigTuple’s solution. An investor in the start-up, Byrapaneni says, “Diagnostics is a large space where AI can help. SigTuple has wisely chosen pathology as their first choice.” According to him, pathology tests are frequent and present a huge global opportunity. 

Tag team 

SigTuple’s founding team, Pandey, Anand and Dastidar, came together in 2012. They were working on a big data project at American Express. After two and a half years, they started looking for more exciting opportunities and healthcare caught their attention. None of the three had any experience in the medical field, but some developments around the time — Google acquiring Deepmind in 2014 and Google Ventures’ announcement about investing one third of their capital in the healthcare sector every year — jacked up their interest. 

So, as an initial attempt, end of 2014, they downloaded a cache of data from pap-smear tests, used to detect cervical cancer, which was available on a Kaggle (a platform for data scientists). They coded a simple algorithm to read this data and detect if the person tested positive or negative for cervical cancer. 

The accuracy wasn’t impressive, but they went ahead and showed it to doctors, who were quite excited to see it. Encouraged by this response, they registered the company in July 2015. 

But they didn’t have adequate use cases or a defined focus area. So they reached out to Anand Diagnostic Laboratory in Bengaluru. With the support of Dr. Sujay Prasad, they started defining the focus area for the company. 

They found that many of the tests were prohibitively expensive, because the devices and reagents used did not come cheap. Their results were also widely variable, since accuracy of manual analysis of slides depended on factors such as medical expertise, skill sets, experience and even the state of mind of the technician. Byrapaneni adds, “No matter how experienced, a pathologist can see only limited slides a day. Also there is no way to cross-check it because, after the pathologist examines it, the slide is discarded. Shortage of pathologists across the country particularly in tier III and IV cities is another problem.” Currently there are close to 19,000 qualified pathologists who are serving more than 300,000 labs and collection centres.

Keeping all these in mind, by September that year, the trio decided they would solve for affordability, accuracy and accessibility. They would only work on high-volume tests which required analysis by a medical expert. “Common tests such as checking blood-sugar levels, can be done at home. We don’t need a pathologist. But, a complete blood-count test requires an expert,” says Pandey. 

Manish Singhal Founding partner, Pi VenturesThey started working on five solutions. Three were in pathology for blood, urine and semen, and two in ophthalmology, around retinal scan. But, they soon realised that the industry had no digital data. To build a data bank, they needed hardware and the team came up with something, which in Pandey’s words, was jugaad. They fixed a holder onto a standard microscope and mounted a mobile phone on it. The phone would click pictures of the slides and scan it. 

It was a quick and an innovative solution, but the phone-based scanner-digitiser wasn’t scalable or sustainable. Pandey explains with an anecdote. A lab owner in Allahabad was anxious that his technicians would steal the smartphone, and wondered if he would additionally need to install surveillance cameras. “This was the first learning. Until you convert it to a medical device, you can’t sell it,” Pandey remembers. In another one and a half years, they converted the ad-hoc device into a powerful scanner. 

This improved device did away with costly reagents used to trigger chemical reactions in a test sample, through which variations in its parameters such as colour, optical density and turbidity are brought out starkly. Variations from the normal values indicate the health or malaise of a sample. The chemical agents act, sort of, like highlighters. SigTuple’s product did digital reading of the sample without the need for a ‘highlighter’. 

The start-up’s product was also compact. What was available in the industry till then were separate devices to test blood, urine and semen, whereas AI 100 was good for all three. The cloud model also cut down on logistics cost for shipping samples. Thus, they solved for affordability and accessibility, and then for accuracy. Since the analysis was done by AI models, and not a tired pathologist stressed out after reviewing multiple slides under a microscope, there was consistent accuracy. 

Trial trail

To create its data set and train its AI model, the company partnered with multiple hospitals and laboratory chains. The partners gave them the data and SigTuple’s panel of pathologists helped them annotate the data. The start-up scanned over 25,000 pathology slides, developing a huge data set (one slide gives hundreds of data points), to train the models. Once the device, dataset and solutions were ready, SigTuple decided to go for clinical trials.

“We conducted more than eight clinical trials and analysed more than 15,000 samples. After we got decent results, when measured against existing devices, we went to Drug Controller General of India (DCGI),” says Pandey. 

SigTuple has started to sell their haematology solution in India, beta trials of their urology test is being done in their Bengaluru reference lab, and the beta version of their retinal-scan product has also been deployed at an eye hospital in Bengaluru. 

To gain credibility and to explore  overseas markets, SigTuple has applied to the US Food and Drug Administration (USFDA) for a licence. The start-up has filed for 21 patents till date and has published more than 18 papers in international and Indian journals.

SigTuple is targeting about 100 installations by the end of FY19. It is targeting exponential growth thereafter with its FY20 target of 600 installations across Indian and overseas markets, fetching a minimum revenue of $5 million. As compared to other devices in the market, the cost savings for a lab will be around 70% on average, claims Pandey.

SigTuple started deployments last April and has completed about 14 so far around Bengaluru and has an order book of more than 20 devices from the city. “We plan to expand into other cities once our manufacturing pipeline is completely operational by first quarter of 2019,” says Pandey. To scale up, SigTuple plans to outsource manufacturing of the hardware, for which they have signed up with a USFDA registered partner. SigTuple is also working on testing the waters in Bangladesh and Indonesia by 2019.

The start-up barely has any rivals in India. But, it is not so globally. Big incumbents such as Roche and Philips could pose serious competition. Roche has an alternative for haematological tests, though it has not been launched in India. It acquired a company called Constitution Medical Investors (CMI) in 2013 with its trademarked Bloodhound technology. In urology tests, there is the Kobe-headquartered Sysmex. Philips is focused on histopathology (study of tissues).

But Pandey is not worried. He says that hardly anyone offers a single solution for all three biological samples, like SigTuple does. The price points, too, are way high at the global players. Devices of the competitors are 5x to 20x more expensive than SigTuple’s, according to Pandey. “The philosophy of companies such as Roche and Philips is to come up with world-class hardware to give you high- quality image. Our philosophy is absolutely different. We believe in using low-cost hardware which can do bare minimum kind of imaging, which is good enough for our AI model to classify and provide the same accuracy as the biggies,” he says. 

Competition from other start-ups such as Morphle, Alexapath and Pathomation is heating up by the day. The SigTuple team tries to stay ahead by calibrating its products to changing requirements. For example, they are making their solutions ready for poorly networked rural India. SigTuple’s present solution requires 3G connectivity and, since that is a challenge in remote locations, the team is designing another version which wouldn’t require cloud connectivity constantly. The entire processing could be done locally within the device itself and 2G connectivity would be enough to transfer the report generated. The product is expected to be ready by end of 2019.

Singhal of Pi Ventures says the start-up is also trying new business models. “It has launched a diagnostics chain called Humain Diagnostics, which is an AI-powered diagnostics lab. So from a B2B player, they are trying a B2C positioning.” However, he notes, there would be challenges to scaling it up and building market adoption. Pandey does not believe that SigTuple’s clients will see Humain Diagnostics, started in January 2018, as competition. “In fact, we are collaborating with some of them like Dr Lal PathLabs, Metropolis and Asian Diagnostics Labs for Humain Diagnostics,” he says. The market for diagnostics is massive too. “None of the incumbents even cover 10% of it,” he adds. 

This side of their business has the same set of challenges faced earlier by their AI platform  — from increasing cost of data acquisition, longer time to take the product to beta stage due to the absence of an in-house lab and lack of a platform on which the team could test their solution in a controlled manner commercially before distributing it to customers. The team has launched three labs in Bengaluru on November 10 and plans to extend it to another 10 cities in the next 18 months. “We will continue to focus on high-volume tests, and also impact lives of patients directly through our B2C positioning,” signs off Pandey.


You don’t want to be left behind. Do you?

Our work is exclusively for discerning readers. To read our edgy stories and access our archives, you’ve to subscribe