Hello, everyone. Good afternoon. I am Ananya Padhiari. I work as a software developer with Arkansas Children's Research Institute. It's been couple of years now.And thank you everybody who is, watching this video or would be watching this video. So today, let me tell something about myself. So I have been a developer for, ten years right now, and I in the IT industry, and I have been working with different in different fields with, finance, with the retail. And this is the health care that I have been open, to working right now. So today, it's a pleasure to be here to speak on a topic that sits on an exciting intersection of AI, cognitive science, and child nutrition. Child nutrition, a space where technology meets humanity. To solve some of our most pressing health and development challenges.

And it's mostly, based on the child nutrition children, like, under the age of two years from the birth. So in the past few decades, we have made remarkable strides in understanding the human brain and the role of nutrition in cognitive development, especially in children. But despite all we have learned, there remain there remain still so many gaps, critical gaps, like early prediction, personalized intervention, how nutrition affects their brain development. That's where AI and machine learning come in, not just for automation, but as powerful lenses for discovery. By leveraging machine learning models, we can analyze vast datasets, from dietary patterns, growth metrics, to brain imaging, and behavioral assessments. These models help to uncover subtle non obvious patterns that traditional statistical models might miss.

More importantly, they allow us to build predictive frameworks identifying children at risk of cognitive delays due to pure pure nutrition, target and invent interventions. So there's a deeper connection. In many ways, the structure of artificial intelligence and neural networks, those are the, foundational architectures in deep learning, was inspired by our understanding of the human brain, which is very complex, part of the body. So today, we will come full circle. We are now using those same AI models to better understand how the real world the real neural networks in children's brain develop and respond to nutrition. This synergy between neuroscience and machine learning offers the potential to revolutionize cognitive science. We can simulate learning processes, decode patterns in neural development, and even personalized nutrition plans that optimize cognitive outcomes for each child based on their unique biology and environment.

It's also important to highlight the ethical and inclusive use of these technologies. As we build models and interventions, we must ensure that they reflect the diversity of children's experiences, the socioeconomic, cultural, genetic, and that our systems promote equity, not bias. So here I present my slide. Decoding the human brain, leveraging AI and machine learning to understand neural networks and advanced and advanced cognitive science in child nutrition. So the human brain, as we know, is very complex to understand. And, as children in their development phase, it's better to understand how the neurons work, how they connect. So we'll connect three big ideas here, the artificial intelligence, the biology of developing brain, and the power of nutrition. The key takeaways from this are AI and machine learning in brain science, how artificial intelligence and machine learning revolutionize our understanding of the brain's neural networks.

Then we have the linking nutrition and brain development. Then we have the advancing personalized nutrition, like how predictive models can help, make strategies about nutrition for each and every different child. Then the real world applications, how nowadays the brain computing interfaces are being utilized to monitor the brain activity and to develop some analysis based on models. Then there comes the ethical considerations as well. So why does it matter, this nutrition child nutrition relating to AI? Because there are 86,000,000 billion neurons, and those are like one fifty trillion synapses. So nutrition plays almost approximately 45% of active I mean, development, which relates to death in children who are less than five years.

And this is very it limits the cognitive potential in millions more we are not aware of. And this AI gives us computational muscle to translate brain scans and diet logs into actionable insight. So from neurons to algorithm, neurons that are that that like the Hebbs rule. Neurons that fire together, wire together. So it's the backbone of today's back propagation. Convolutional layers imitate the virtual cortex. Now the loop has reversed. AI is coming back to help us understand the real neurons. Here, the key idea is the brain and AI are now in a feedback loop. Each inspires the other. The AI tools in neuroscience. So there are three main pillars, the neuroimaging analytics, the con connectomatic connectomics, sorry, simulation, and digital twins.

So the neuroimaging analytics is the deep CNNs clean and segment massive fMRI or diffusion delta datasets revealing hidden microstructures. So this basically means that whatever we do the MRI scanning and the connections, The CNN refers to the connections that each neuron gives a spark to other one, and that connection, maybe it is not straight, but, in a different way. So those all are being captured. So these are like hidden microstructures. Now connectomics. Connectomics. It is a graphical neuron network that connects these wires, the the different neurons wiring to each other into diagrams so that we can link them and we can understand a child's IQ or age ADHD or something else. So these are the connectome the connectomex. Now the simulation and digital twins. It's a reinforcement learning agent like, the brain circuits letting us test how the plasticity responds to injury or how each how it reacts to different emotions, different environment situations.

So these are the three important tools, AI tools in neuroscience. Now there has been breakthrough in computer, brain computer interfaces. Like in 2023, a transformer model converted electro corticography into synthetic speech at 62 words per minute, restoring communication to a logged in patient. So electro corticography is actually a method of, where an MRI scan, scans the brain. Then this electrocorticography helps this, capture the images and the speed of, how fast a brain reacts to certain activity, like a logged in patient in the sense the patient is paralyzed, but their brain is still functional. So how does it capture and, how many words per minute restrained communication? That has been a big, breakthrough. Then similar models let people with spinal injuries move a cursor with a latency under ninety milliseconds. We have seen such examples, in in today's world.

So these, are not just the language in, but these are the neural spikes that is more powerful, I think, in the medical world where we can help human beings who are having restricted, communication with with people around them to help them come into and mingle with everybody as normal as they can.

The first one thousand days. So what do we observe here? Let's go back to the childhood from the cons from the day they are, the conception happens to the age of two years. It's about eighty percent of cortical architecture. Then projects like baby connectome are scanning infants every six months and logging everything they eat. That's terabytes of data begging for machine learning. So we have been logging so many information data about child's nutrition, about their development. So there's a lot of data out there which needs analysis and which needs to be structured. So there are lots of information that is useful and can be made helpful to people, to researchers, to doctors, to parents to to come and see what actually nutrition does develop, helping developing the brain activity.

So nutrients that wire the brain, we have mainly for the brain, we have, DHA, which we might have heard a lot about, then, zinc, then we do have dopamine. And for these are the main ingredients which needs to be in the nutrition for the brain development. So we can, have a model in the way that they can predict a child's deficiency or what kind of nutrients the child needs in order to fulfill these DHA, zinc, dopamine, iron, in their, everyday nutrition for chi for their development. And let's go to a case study for, different nutrients. First, let's see the iron. So there are four six hundred healthy toddlers, roughly 18 to 36 old. The key measurements for each child, serum ferritin, the body's main iron storage protein drawn from a small blood sample, then the resting state fMRI to ensure to measure how strongly different brain regions in the d n DMN default mode network talk to one another. So this DMN is a circuit that helps that supports self reflection, daydreaming, con their everything cognitive skills. So this is one key measurement. Now what is the analytic approach?

What can we do with this data? So preprocessed fMRI can be used to extract single metric of DMM efficiency. Now how like, how smoothly one, neuron can, attach to others. So it's like one single flow. It's not complicated and a zigzag one. So it's a basic line we can draw. Then a train, gradient boosting regression model. This is a machine learning model we can implement to predict the efficiency from the serum that is drawn while controlling different criterias like age, sex, still emotion, socioeconomic status. So we can draw a gradient depending on different filters, I would say here. Then why is gradient boosting regression model important? First thing, it's, nonlinear. It doesn't have a it it takes the nonlinear relationship. It doesn't have to be a straight line relationship all the time.

And there are you can see the sharp drop offs and, drop offs of plateaus nonlinear effects. But the one side effect major, I think, has been the noise would be a big factor which, impacts the analysis here with the regression model. Now let's come to DHA, another, key factor for brain development. It's a key omega three that fat that builds neural neural membranes, especially in the last trimester when the child circuits are wiring up. So it's like the numbers, approximately two hundred milligram per day, boots a toddler can say 50 words at 18. So that's a quite sizable leap for any child. Why it's convincing?

Because statistically, when we compare the family's income, the parental education, then how much we interact with the children, compared to that, the nutrition has been showing a significant impact on the brain development of child. So what now? So we had do have the prenatal, supplements or DHA fortified foods that can help in early development of the brain of child from the from age zero. But the point being, we can have more of the different sub, supplements in terms of food as well, not just, our medicinal supplement. Now the personalized nutrition. What do we mean by personalized nutrition? So machine learning and artificial intelligence can help with the, with the, let's take a child's brain as a puzzle. So this puzzle can be saw cannot I'm not saying it's it it is solvable, but we can check what, are the ingredients that how we can connect them, like the DNA, microbiome, MRI scans, and their dialogues.

So these factors can help us in building a model which can help understand what nutrition can be, or a personalized nutrition for each child can be made so that their brain development is, in a proper, rate, in a progressive manner. So we can have, like, what will it benefit to the parents or the clinicians, like a app style nutrient prescription card, which fits cultural preferences. Also, it can give alerts like what, prompting the diet tweaks to the, parents or clinicians who are monitoring their nutrition on brain development. Real world testing. So there has been, trials now running in Texas and Singapore to see if AI guided diets outperform standard guidelines. So there has been studies, research going on as well. The place where I work there also has is, ongoing studies about brain development. So we are seeing lot of AI being implemented in child nutrition nowadays to help them get personalized nutrition, at the top of their fingers to parents, to clinicians, to doctors, to, I think, to people out there, to everybody.

So we are moving from, like, one size fits all to optimize each unique wiring. Then the implementation road. There can be so many shared data cloud biomarkers, explanation dashboards. We can have dashboards for different, nutrients and see how it is going for everybody so that it's we can we need to be as transparent as possible. Then the ethical and social guidelines. So there is privacy risk. Like, if we, get the data of brain scans of a patient or a child, we need the biometric IDs as well. These need to be secured, in a very secured system so that there's no data leak. Then they bias in AI. So AI is mostly what we feed. Like, as a human being, whatever datasets we feed into the system, that's the result it's going to predict us.

It is based mostly not only, we should provide a particular, group of data, but diversity. We should look at that. Then mind reading concerns. So nowadays, we do have so many advanced technologies that are reading the minds. So that might be a factor we need to consider while creating this nutrition model or analysis on this. What solutions do we have? We can have private built in from the start mandate training on diverse population then regulates, regulations like the upcoming US neurotech ethics Act to guide safe ethical use. We can always implement such ethical, guidelines. Then we do have governance check rules, the model card sunset clauses. Bottom line is we should make sure data use is ethical, transparent, and that the it respects child's privacy while enabling impactful research.

Then we do have future horizons, like they can wear the head headbands or, the BCI like a closed loop brain computer interface, then AI design food metrics can also help them. The key message being it the future holds exciting tech that merges nutrition and brain science in real time, radically transforming how we support child cognition and development. That is all I wanted to share today. Thank you for your time today.