Updated: Apr 16
DeepTek, a PUNE AI-enabled radiology platform, is playing an important role in the precise diagnosis of diseases such as tuberculosis and Covid-19.
So far, the Pune-based startup has raised strategic investment from a number of investors, and it plans to raise another round of funding within the next six months.
DeepTek was founded in 2017 by Ajit Patil and Amit Kharat. Patil graduated from SSPMS school and COEP engineering school in 1992. He holds a Master's degree in Industrial Engineering and Operations Research from IIT-Kharagpur. Amit Kharat, who holds a DNB and a PhD in Radiology, has worked in radiology for the past 17 years.
Ajit Patil's family is from the Sangli district, and his father, Surgonda Patil, used to work as an engineer for Kirloskar's. Patil's father had founded a manufacturing workshop that he managed after hours.
Patil says, “My father used to return around 6 pm from office and pick up his bag and go to the workshop. He was a role model for me to become an entrepreneur. I was a very shy person and wasn’t sure whether I was geared to be a good entrepreneur.”
“Most students from COEP at the time we're planning to study or work in the US, but I wanted to do business in India. I worked for one year in Telco under the sandwich programme, but then I realised and decided that I am not going to do a job anymore. I appeared for the GATE exam and went to IIT Kharagpur in 1992. During one of the breaks, I had come to Pune and met a friend from COEP. He had done his mechanical engineering but was working at a software company called Fujitsu. I was surprised. When I went back to IIT, I was doing an industrial management programme and was asked to analyse the opportunities for the software industry in India. That time, in 1994, I realised that the software industry is a huge wave coming and can transform India. So, I also joined Fujitsu to gain some experience in the software industry. One fine day I decided to stop and go back to India. I met my father and told him I want to do something in software. My father was in mechanical engineering, but he encouraged me. That gave me confidence. I started the company Vertex Software, a Japan-focused software services firm that had strategic equity investment from Mitsui and exited to NTT DATA.
“We were one of the largest Indian companies working on the Japanese market, with over 700 employees, of which 200 spoke Japanese.”
What exactly is radiology AI?
AI is extremely strong in image analytics, and radiology is all about images from MRIs, CT scans, and X-rays. Radiology AI is the process of analysing these images and determining the pathology or disease, as well as the degree to which it has affected or advanced.
Each computer and its settings are distinct, as is the human body. It is not an easy task to develop an AI solution that works across these machines and body types. Furthermore, radiologists are in short supply in almost every country. Owing to regulatory concerns, India has only 8,000 radiologists who perform CT scans and MRIs, and another 8,000 who perform sonography. These are small numbers as compared to India's population.
Ajit Patil left Vertex in 2013 to pursue other opportunities as an entrepreneur. Patil remembered“In 2017, I was talking to Tsyoshi Kitani, then CTO of NTT Data. He mentioned that they were interested in radiology AI. They had a product in the US which was used in 1,000 hospitals there. In AI, data is the oil and without data you cannot do AI. Kitani said they had the data and wanted to explore something in this space and if I could do something, they will look at it. I started researching. I met almost 30 to 40 people including industry experts, educational experts, AI experts, and radiologists.”
“I met Dr Amit Kharat and offered him co-founder of the start-up. Aniruddha Pant, who is a data scientist and also an alumnus of COEP, also came along with us at that time. We had gone to US to visit an exhibition related to radiology, organised by the Radiology Society of America. We were there for four-five days. We talked to lot of companies doing AI and tried to understand what they were doing. We learnt that none of the companies – including a few which were funded – had any commercial adoption. We were on the way back to Newark airport and almost decided that we will not pursue this idea. After a while, we realised that this idea would work only if we take radiology experts on board. We should not just depend on AI, but ask the experts to take ownership. This “expert in loop” model clicked and we all were energised. We thought this is where we can create and deliver value to the hospital,” Patil explained.
Patil says, “In 2017, we started working on the model. We did a few things initially and then we approached NTT DATA again. NTT DATA took an immediate interest and decided to invest in our company in 2018. One of the earlier investors, Mitsui also expressed their interest and their group company Nobori, Japan’s largest Cloud PACS provider, came on board as a strategic investor. Aniruddha Pant helped us develop the data science experts’ team. We have a total team of 70 people, including 20 radiologists and 40 data science and technology experts, along with a strong leadership team. This combination is a unique thing we have created so far.”
Strategy for Commercial Adoption
“Commercial adoption of radiology AI workflow is not happening even now. First intervention in any hospital is an X-ray, but most of them go unreported. DeepTek’s solution is very powerful on X-rays as it can automatically identify 20 pathologies. DeepTek believes in augmenting the radiologists and not replacing them,” he stated.
Says Patil, “With the “expert in the loop model” we managed to get 120 hospitals and imaging centres as our paying customers. These are mostly small and medium establishments situated in Maharashtra. Amit Kharat’s teleradiology customers became our first customers. We migrated those customers on our AI platform and now we are adding 10 customers every month. Last year, due to Covid and the consequent lockdown, we explored the overseas opportunities and got customers in Japan. We are conducting pilots in Indonesia, Phillipines and Africa also. NTT DATA and Nobori helped us with relationships and connections in getting the international customers. Apart from NTT DATA and Nobori, Gurugram-based Indo-Japanese Venture Accelerator GHV and Pune-based Pentathlon Ventures has invested in DeepTek.”
TB screening method
DeepTek has collaborated with the Greater Chennai Corporation (GCC) to provide an artificial intelligence-enabled solution for tuberculosis screening and triage using automated chest X-rays. Since 2019, DeepTek has screened 100,000 people using its AI-first population screening tool, GENKI (Japanese for healthy). This current collaboration has been expanded to screen 150,000 patients by 2022.
Covid-19 detection tool that works
A CT scan has developed into a responsive and efficient method for diagnosing Covid-19 infection. Patil elaborated on this approach, “There are certain kind of patches in the CT scan images which are indicative of Covid infection. CT scan has become one of the primary techniques for segregation of Covid patients in the hospitals. A lot of research has been done globally in Japan, China, Singapore, regarding this. DeepTek has also conducted research in collaboration with Nanavati hospital and Ruby Hall Clinic (RHC). We have done a ‘proof of concept’ with Nanavati hospital and we are in the process of publishing papers related to the research conducted with RHC.”
Tsyoshi Kitani, president and CEO of NTT DATA INTELLILINK Corporation, said “I am pleased to see the increasing maturity of AI technology and medical imaging services at DeepTek since the beginning of its business in 2018. The technology and service helps to solve society issues as we witness in the Chennai TB elimination project. NTT DATA Intellilink will continue working with DeepTek in developing emerging technology and contributing to society,”.
To help their work, Newsmusk allows writers to use primary sources. White papers, government data, initial reporting, and interviews with industry experts are only a few examples. Where relevant, we also cite original research from other respected publishers.
Source - Hindustan Times