Artificial Intelligence (AI) and Machine Learning (ML) are key topics of discussion today and they will have a big impact on industries across sectors.
This is an edited excerpt from an interview of Aniruddha Pant, the founder and Chief Data Scientist of DeepTek Inc. conducted by Amit Paranjape, Chairman of the IT & ITES Committee at MCCIA, as a part of MCCIA’s YouTube series, ‘Decoding Artificial Intelligence with Industry Leaders’. Dr. Pant talks about the applications of AI in healthcare, the use of data in medical imaging analysis, how AI and ML can be used to predict symptoms of a patient, and the future of the healthcare industry in terms of AI.
Dr. Pant completed his mechanical engineering from the College of Engineering, Pune in 1994 and then went on to do his Ph.D. in Control Systems from the University of Berkeley. Subsequently, he worked on statistical learning theory as a post-doctoral researcher at UC Berkeley where they created an algorithm for support vector machines that would use past data to identify patterns to optimize decision making.
While talking about his entrepreneurial journey, he explains that after working for close to 6 years in the industry, he decided to take the plunge and start his own venture in 2009. That was how AlgoAnalytics was born, a firm where they applied ML to financial markets for derivative forecasting. They worked with high net-worth individuals (HNIs) and proprietary desks and made a reasonable amount of money for 3–4 years. However, in 2015, it became difficult to sustain the company’s cash flows solely on this and so they decided to get into services as well where they bagged many outsourcing contracts to work on the core algorithm of new projects. He says that as they were the pioneers in the fintech space, the biggest hurdle they faced was to work in the absence of a regulatory framework.
He further speaks about his experience of data mining in the retail sector. While working for a large US-based retailer, he realized that it is challenging to develop models from data that is ‘dirty’ or inconsistent as it is in different formats and unstructured. But after refining, the data can be used to fine-tune the recommendation algorithms of the retailer. Several factors affect the algorithms and by designing a system that will combine these factors, the algorithms can be automated. While AutoML is a new field, he thinks that it will boom in 2020–21. He also says that an Internet of Things (IoT) revolution is taking place in the manufacturing sector However, there are a lot of friction factors in this sector and it will be tough to get the people within the organization to accept the use of AI in the company.
He then moves on to talk about DeepTek, a start-up he co-founded which applies AI in the field of medical imaging. At DeepTek, they have developed algorithms that look at x-rays, identify, and segment the image and create a standard radiological report. That output is fed into a natural processing neural network, which generates the report that appears as though it is written by a human being. The biggest advantage of this is that it saves a lot of time. These algorithms have been used by the Greater Chennai Corporation for more than a year and it has led to them diagnosing more than 1000 patients per month. The team at DeepTek has also adapted its pneumonia detection algorithm to test for Covid-19. They are trying to use the Covid-19 algorithm on x-rays and CT scans and combine it with other clinical parameters like the SpO2 reading. In his opinion, 2 factors differentiate DeepTek from its competitors. The first one is the development of a suite of algorithms for a single body part as opposed to only a single algorithm for a disease. For example, it is better to have a set of algorithms detecting 95% of the chest pathologies than having a single one for pneumonia. The second factor is having highly qualified and experienced radiologists to supervise the process and correct the algorithm if it goes wrong. However, he believes that the algorithm shouldn’t completely be automated right now because if something goes wrong at the endpoint, the algorithm wouldn’t be able to detect the change and keep predicting things on the same data which could prove to be dangerous. He also explains that while collecting data for training they use privacy reserved machine learning algorithms that don’t take possession of the data but install some models and take only that information in at the server and then train the models using information from different sources. He also says that small data machine learning is another area that DeepTek is looking at.
When asked about the Indian talent in the space of AI, he says that the freshers have a lot of potential and he believes in hiring them as they bring a lot of enthusiasm to the table and are interested in trying out new ideas.
He concludes by saying that by 2030, the healthcare sector will undergo a sea change. Physicians will be assisted by AI and the way diagnosis takes place will be revolutionized.
(You can watch the full interview at https://www.youtube.com/watch?v=Ik9ayHTWS10 )