Automating Microscopy to Deliver Quicker, Better Diagnostics to Patients
An Indian company is automating microscopy in medical labs with an artificial intelligence-powered solution, improving efficiency, and speed, and helping doctors quickly detect critical diseases.
In a medical laboratory, most tests are automated, except for microscopic examination, which still relies on human visual evaluation; a slow process subject to errors, inefficiency, and affected by the growing scarcity of medical personnel. SigTuple’s system pre-classifies cells and provides a report in minutes.
Tathagato Rai Dastidar, best known as “Tatha”, had trouble deciding what he liked the most when he was in high school: physics or computer science. He chose the latter and after a B.Tech., went on to work for the national semi-conductor sector writing software for the automation of hardware design, while completing his Ph.D. in computer science at the same time.
Seven years and a failed attempt at entrepreneurship later, Tathagato entered the internet domain.
He joined Yahoo and helped with search functions, getting familiar with big data, image, and video analysis. He later went to work with American Express and built up the company’s big data capabilities. “Amex,” he said, “was one of the first financial institutions to jump to the big data world.”
By 2014-2015, “when AI was finally coming of age and we could actually see results which were paralleling human intelligence,” Tathagato and two former colleagues from American Express, Rohit Pandey and Apurv Anand, started exploring AI applications. “The idea was to build something that would make a difference in people’s lives,” he recalled. No ideas rang the right bell until they considered healthcare.
Neither Tathagato nor Rohit or Apurv had any medical background “but we knew a lot of data in healthcare could be used to make things better.” “It was a little bit of a leap of faith for us,” he said. Fortunately, they found advice from “visionary medical professionals, doctors, and pathologists” in Bengaluru, where the company is headquartered, and were provided with the initial set of data to get started on their project in 2015.
AI Image Analysis with Diagnosis left to Humans
The three co-founders focused on the automation of microscopy. The current manual process “is inefficient, error-prone because it depends on external factors, such as doctors’ fatigue level and somewhat inaccessible because the pathologist, the microscope, and the sample all have to be at the same place, at the same time,” Tathagato explained.
SigTuple provides a combination of hardware and software solutions. The hardware is an automated digital microscope (AI100), which can convert any medical specimen into microscopic digital images and upload them to a cloud platform. Two AI-powered software: Shrava (for urine) and Shonit (for blood), analyze those images, and pre-classifies cells into types and sub-types, delivering results in minutes. “Our system is better at picking up rarer abnormalities than a human pathologist,” Tathagato underlined.
The system delivers clinical insights, which are available to pathologists on a standard web browser. The pathologists could be anywhere, he said, adding that the clinical insights are backed-up with visual evidence so that the pathologist can corroborate the AI-defined insights. “Some cells have a borderline aspect and the pathologist may agree or disagree with our classification so he/she can change the classification to edit the results.” The system also allows for collaborative review between colleagues without having to ship samples. “Now the pathologist just sends a link”, he said.
“We can predict certain pathologies, such as blood cancer or anemia, but we stop short of doing that because ultimately we want the end diagnosis to come from a human, not from a machine.”
The SigTuple system performs all analyses normally conducted on blood samples, including differential counts, and red cell and platelets morphology analysis. Health issues that can be detected include blood cancer, such as leukemia, infections, and inflammations, different kinds of anemia, as well as viral infections such as Dengue, and hemoparasite affections, like malaria.
“When it comes to urine, a normal urine sample should have nothing in it,” Tathagato explained, adding that cells in urine are always a sign of health issues, such as a urinary tract infection, or internal trauma.
The SigTuple’s system does not require any personally identifiable information, or demographic information about patients, explained Tathagato. “All we require is a barcode/ID provided by the user, either a laboratory or a hospital, and we cannot tie the barcode to the patient.” The cloud platform is strictly access controlled and the company is compliant with the European Union General Data Protection Regulation (GDPR) and the United States Health Insurance Portability and Accountability Act of 1996 (HIPAA).
The ultimate aim of SigTuple is to sell its hardware. The company then collects service fees for its software. However, in certain disadvantaged settings, they rent the microscope.
SigTuple’s customers are laboratories, either in hospitals or in stand-alone laboratories.
The active commercialization of the SigTuple system started in the fall of 2021. The samples of “a few hundred patients” are analyzed every day all over India. Since the commercialization started, thirty units have been installed in the country.
With 19 active patents, six of which are also registered in the United States, its logo and name protected, SigTuple understands the value of IP assets.
“The way we look at IP is in two ways,” Tathagato said. The first is to give the company a defensive mechanism and avoid being sued over a technology that we have patented, he explained.
The other serves as a strong motivational factor for the company’s employees. “Getting a patent to your name is a huge motivational factor for many young engineers,” he said, adding the company “has a good reward system for IP.
SigTuple, with a staff of 70, is now in full throttle as it seeks to multiply the number of its AI100 Microscopes in India and expand to the international market. The company is currently in the process of filing for the US Food and Drug Administration’s approval to enter the US market and is also targeting the European, Southeast Asian, and Middle East markets. The system is affordable for most Indian laboratories, which makes it affordable anywhere else in the world, Tathagato noted, underlining that fact as being a serious competitive advantage.
At the same time, the company is busy with R&D to produce a multiple-use device for laboratories, which could handle more tests from the same hardware. This could include microbiology, pap smears (cervical cancer screening), and biopsies.
SigTuple plans to set up “points of care” devices that can perform basic testing outside of traditional laboratories, such as emergency rooms, or even waiting rooms in hospitals and clinics so doctors can access the information they need to prescribe a medication for example.
To circumvent the absence of a lab technician to prepare sample slides, the company is working on a device using microfluidics technology. Microfluidics is a branch of science, which allows the manipulation of very small quantities of liquid through narrow channels, Tathagato explained. The samples can be prepared very easily and put into the microscopic reader with that technology, he said, adding that the product could be out in the next two years.