Good day, Mrinalini Bhagawati. I'm Arjun. I've had a look through your background and your professional record. I'm interested in hearing more about your work and how you see yourself fitting in here with our faculty. Ready to start?

Yeah.

Could you briefly share your professional journey in academia?

Uh, yeah. Good morning.

I'm I'm Doctor Manalini Bhagwati. I have completed my PhD in Biomedical Engineering with a domain in AI in healthcare.

Of in.

From a Northeastern Hill university in the Department of Biomedical Engineering.

By in last last year in December, December 2025 and before that I have completed my M Tech in Computational Biology from Pondicherry University.

In 2020 and before that I have done my Btech in biomedical engineering.

In from Northeastern Hill University in 2017 and during and I also have a research background in the domain of AI in healthcare with having with having 14 journal papers and.

Other publications like book chapters and conference. I have 21 publications in total.

Including other.

And I also hold the experience of two years and three months in teaching as a.

As a as a guest faculty.

In the Department of Biomedical Engineering, Northeastern Hill University.

That, that's all, is my description in academia.

Thank you for sharing that, Professor Bhagwati.

Thank you.

Professor, let's begin by verifying your expertise. Could you explain how you have applied artificial intelligence or machine learning in healthcare-related research or projects? Specifically, I'd like to hear about the methodologies and outcomes from your work.

OK so in my during my research in in PhD I have applied artificial intelligence method especially the deep learning methods like RNN, LSTM and Gru.

In for the risk prediction in the cardiovascular disease disease. So they they're basically we have taken the.

Uh, risk factors from, uh, associated with the cardiovascular disease and the ground truth taken where the angiographic scores and then we have, uh, then, uh, we have applied that to the deep learning methods and then then got the different parameters.

Like different parameters as in since the evolution the parameters like accuracy, sensitivity, specificity. Like that and.

In accordance with the deep learning methods, we have also taken the bidirectional deep learning algorithms and to further enhance the accuracy we have done, we have added the attention.

Module or attention block to the this this three models that I have used that were RNN, Gru and LSTM and in accordance with the bidirectional methods for these three and we have further to enhance the accuracy we have. Done the done experiment in the hybrid mode also.

Where we have.

Edit 2 Algorithms.

In both bidirectional and unidirectional paradigm. So we have uh we have added the a error in with Gru.

Then a Gru with LSTM and RNN with LSTM like that. So in this way we have made.

33 combinations in both bidirectional and unidirectional DL paradigms.

And after that we have.

Taken the performance evaluation parameters and also have plotted the Roc curves.

For each.

Each method and finally in conclusion we got the result that the attention based.

Hybrid DL modules are the best performing modules in terms of accuracy and other parameters.

So that was the result that we had got and we inter we have interpreted that the attention modules are better and also secondly we have interpreted that bidirectional paradigms are are good when compared to the unidirectional deal paradigms as.

As they are giving higher accuracy.

Yeah, and.

Yeah, thank you. That, uh, describes my uh.

Experiment, uh, tell protocol and the methodology in, especially the methodology part in my research.

Which include AI in healthcare.

Thank you, Professor. Your application of deep learning and attention modules for cardiovascular disease risk prediction demonstrates a sophisticated approach. Could you elaborate on how you addressed potential biases in data selection or preprocessing during this research? Specifically, how did you ensure your models were robust across diverse patient populations?

Uh, yeah. So, uh, for uh, addressing the bias we have done, we have used the different K protocols like K2 from starting from K2 protocol, K protocol are basically the cross validation protocols.

Like K2K4K5 and K10 and also TT model is there so to address the bias we have.

Done. Now validation in in the using the K protocol. So we have used K2 then K four then K5 and K10 so as as. As we should get with the increasing and with the increasing size in the.

In the training, training data data increasing size in the training data, our accuracy is increasing with each K model like with with each K protocol. Now that is one aspect where we that have we have used to address the bias in the research and secondly.

To address the different population that you have asked, so we have uh.

Applied the validation through the unseen protocol that is like we have trained our system in one one population and we have tested in the another population which was not the part of the training population. So that means we have tested our model.

In the. With the unseen database that which was not included in the.

Umm, testing a data, uh, in not including the training data set. So uh, and yeah and we have our, our, uh, the data that we have used basically includes multi ethnicity data basically so that can address the. Like populations of different ethnic groups. And in that way we have we can justify.

The use of different populations with different ethnicity like our model is performing good name that type of data sets so that way we can justify the our models.

Yeah, thank.

Understood, Professor. It's clear that your use of cross-validation and multi-ethnic datasets is well-considered to minimize bias and ensure diverse applicability. Let’s proceed. Could you describe your approach to teaching theory and laboratory courses? Specifically, how do you ensure complex concepts are comprehensible to students with varied academic backgrounds?

Uh, yeah. So I am comfortable with the laboratory courses, uh, including, uh, uh, biomedical instrumentation that will include different instruments like ACG, EMG and like that several, uh, instruments in that course I'm comfortable with and. Another course we can uh, have is like AI AI laboratory setup for the.

For the students to get familiar with the AI systems basically in the healthcare field and.

And and and describing about the approach. Firstly, we can describe students about the biological concept so as they get more interested into the course and we can demonstrate them with different simulators that were used in the. Lab and describe them the use of. For the different instrumentation and the real life example we can give them we can also and for in the in the field of AI we can show them different.

Initially different small models so that they can understand and can correlate them with a real life example and the simulators present in the lab even we can.

Develop some model in the MATLAB using the simulators present in the MATLAB module.

And we can, and then later on we can also have a.

Hospital visit to ensure to show the students the different instruments that are there and correlate them with the life example.

Like real real time examples and from different and this will help from the students from different backgrounds to correlate as if we take the simulators.

And also have a visit hospital visit so they can.

Like they can have a real life example and I and I I think with this we can make them more interested or prone to.

Come in the research in the in this particular field of AI in healthcare.

Yeah. Thank you.

Thank you, Professor. Your method of blending theoretical instruction with practical demonstrations and real-life exposure, such as hospital visits, seems well-structured and engaging. For evaluation purposes, how do you ensure your exam formats effectively assess both theoretical understanding and practical skills?

Uh, yeah. For the exam point of view, we can have a further theoretical purpose. We can have some exams that are having the multiple choice questions and initially and so that we can make sure that their concepts are clear. Rather than having a long descriptive.

Question are the examination pattern in the initial part and after that we can go further descriptive type papers questions for the theoretical approach. But I think As for my point of view, we can also have a multiple choice question patterns. So with that we can we will be able to assess more in the concept.

Wise.

One system, if the students are clear or not and for the practical purpose, we can after we have demonstrated the practicals during, we can have practical examinations where we can ask the students to perform the practical in and show and show us the whole methodology and the results that they have acquired. By performing by them.

Themselves by their own, actually using the instruments and the different.

Different models like AI model and.

During the exams. Thank you.

Thank you, Professor. Your examination strategy—balancing multiple-choice questions for concept assessment and hands-on practical evaluations—demonstrates thoughtful consideration of both theoretical understanding and applied skills. Moving forward: guiding student projects and research is critical. Can you share your approach to mentoring students on independent research, particularly in areas like artificial intelligence or health informatics?

Uh, yeah, sure. So, uh, to motivate, uh, other students in the, for the research field, especially in the AI, AI in healthcare, we can, uh, have.

A module included in our curriculum or system. Uh, or the course which will have some short, umm, initially some short courses.

Oh, demonstrating different AI models, both including DL and ML models and they they let them grabs the concept initial concept by taking that courses.

As we can give them some small project for like with having a duration of one to two months side by side with their classes, they can carry on that small project, you know where we can mentor them describing the objectives that they need to achieve from the from the.

And then we can.

Project and and then we can.

See how much they can.

Get get the objective done done in the in the short term projects and like that we can motivate them further.

To make that project in a large scale.

Ways to?

And then there will be more into the research. By taking a small project, they can learn more.

How we can start a research based work basically?

And we can also have them learn how to dig into the literature.

Surface or the literature reviews that we get and how to use the publicly available databases and also different sources where we get the research papers or review papers, basically the articles.

But, uh, and we can, uh, we can, uh.

We can also.

Teach like show them how we can easily get that different articles and then you can show them how the methods to read the lengthy papers basically in a short time by giving them a strategy like we can make point wise system rather than. Point wise documentation rather than reading.

In the hole in the manuscripts. So in that way we can motivate them to work in the research as because they get demotivated when they see the long, lengthy.

Research papers or articles. But we can ask. We can reach them in the ways in which we can. In short time you can read them by making notes like by making points.

Uh, like just firstly focus on lifestyle, but then slowly move to the methodology. If we get, uh, the abstract is, is from abstract. If we see this paper is useful for our research, then we can move to the methods, then to the results. Like that way, some, some sort of different ways we can teach them and get them motivated to, uh, towards the research field.

Thank you.

Professor, communication and structured teaching approaches are crucial for student engagement. Can you describe how you ensure clarity, structure, and accessibility in your lectures or presentations? How do you adapt your approach for students who might struggle with more complex materials?

Yeah, during the lectures we can like, I prefer like firstly giving the, uh, showing them some.

Presentations.

And and making it in a point wise manner or including figures, more number of figures so that they can get engaged in in the lecture mode and and we we can have different different assignments or the OR different small task that.

We can give to the students to, uh, perform like, like we can ask them to make, uh, make group of three, three students and then discuss a particular topic among them. Then, then the group can exchange each of their ideas, uh, that were different ideas that they have gained from the lectures. And then they can discuss the doubts that they have from the lecture.

And.

He uh, in that way, uh, good communication can be developed among the children's and the, among the students and the student and the professor.

Uh, firstly among the groups, then the within the groups, then uh, different two groups can discuss and then.

Understood, Professor. Your emphasis on interactive discussions and group tasks fosters collaborative learning and ensures clarity. Let me move on to verify your academic qualifications and research contributions. You hold a Ph.D. in Biomedical Engineering—could you briefly describe the focus of your doctoral research, particularly its significance within the field?

Uh, yes, my uh, doctoral research was focused in, uh, in the AI in healthcare, uh, domain where like basically I have uh, uh, my topic was to develop a non invasive uh. Cardiff Vascular risk stratification of techniques using AI so.

The where we I have used different deep learning the the different like I have developed a different deep learning paradigms and compared them with the ML methods and my results shows like DL paradigms are.

Are more superior when compared to ML and also I have done the addition of a hybrid mechanism and the attention based mechanism UH to the traditionally based uh DL methods.

And this and.

Yeah. And also the in the in, in, in the factor of risk fact risk variables or factors I have added.

To new types of risk variables that were the Q self means image based uh.

Venue types and also the.

Your medication used, uh phenotypes are added.

Making the the risk factors.

In in four different classes in which first one is the office based level office space physics biomarkers and second one is laboratory based risk biomarkers and 3rd and 4th 1 is the novelty in my research. That is the inclusive usage is the image phenotypes biomarker and.

Made use. Basically we have named it as a risk factors factor group.

The last one is medication used.

Thank you, Professor. Your integration of novel risk variables and advanced deep learning methodologies in your doctoral research demonstrates significant innovation in AI applications for cardiovascular health. Now, could you provide an overview of your research publications in reputed journals, including any notable contributions or impacts these publications have had on the field?

Our scheduled time for the interview is now up. Before we conclude, do you have any questions related to the role or the company?

Uh, yeah. I can, uh, conclude my research, uh, uh, research contribution as I have, uh, I have in total 21, uh, publications, including different general papers and conference paper and book chapters.

Book Chapters.

So I have.

4 research journal papers that were published in different journals like IEEE. IEEE Access and.

Another some few journals where come where that BSPC that is biomedical signal processing journal and.

Another journal is CDT, that is cardiovascular disease and theory, and I have one book chapters here that is that is. Present that got published in the Elsevier book and by the Elizabeth publication and I have different international and international conference present presentations.

Completed like I have presented recently in IT Kharagpur and also presented in Nimhams Bangalore and also had a presentation in in Naipaul, Guwahati like that I have several.

Presentations oral presentation made and I have a few few uh, review papers also like as in number if I say I have total.

UH-21 Publications.

Think.

You’ve provided remarkable detail about your research contributions. If you’re ready to conclude, let me know, Professor. Should we end the interview?

Yeah, I am clear from my side. I have delivered, uh, my expertise and everything, and I would like to thank the VIT team for giving me this opportunity with the interview.

Thank you, Professor, for your thoughtful responses and detailed explanations. Just to confirm, would you like me to formally conclude the interview now?

Uh, yeah, sure. Thank you.