An 11th grader brings technology to canine care
Anshul Bhatt’s non-invasive gait monitoring device detects orthopaedic diseases in dogs
During a trek last year, Anshul Bhatt realised that the family’s eleven-and-a-half-year-old Labrador, Max, was in severe discomfort. While he would normally be running about in the wild, never refusing a dip in a pool of water, he now sat licking his paws in pain. A visit to the vet revealed late-stage arthritis.
“Since it wasn’t diagnosed in time, the treatment is less effective. You can alleviate the pain and make dietary changes, but there isn’t much you can do beyond a point," says Bhatt, 16, a class XI student at Dhirubhai Ambani International School, Mumbai.
The incident got him thinking. By the end of the year, he arrived at the design for PawPath, a non-invasive gait monitoring device that detects orthopaedic and neurodegenerative diseases in canines. It won him second position in the Animal Sciences category at the Regeneron International Science and Engineering Fair Awards ( for students) in the US in May.
During the early days, Bhatt reached out to vets to understand the current methods being used to study different ailments based on dog gait. Visual clues are the most basic form of detection where a vet physically inspects a dog. The other two require an expensive indoor setup—while kinetic gait analysis studies movement using force plates, kinematic analysis utilises a high-end camera to capture a dog’s movement, usually on a treadmill. This would typically be followed by an X-ray while sedated to detect various abnormalities from orthopaedic and neurodegenerative disorders to cruciate ligament tears, osteoarthritis and ataxia.
“From my research, I realised that dogs exhibit altered movement patterns when they are in a lab-based setup since it is not their natural environment. As a result, it may look like they have an issue when there may be none. I started looking at ways to eliminate the subjectivity of visual analysis and to also do it in a way where the dog was comfortable, where the entire process was cost-effective," Bhatt says.
At the heart of his design are inertial measurement unit (IMU) sensors, which are strapped on to each limb to capture motion. The first prototype, which took two months to build, featured a central module that gathered data from the four IMUs and transmitted it via Bluetooth. But besides the weight of the entire system, Bhatt realised that the wires were hampering the dog’s mobility, both resulting in altered movement patterns.
So, he went wireless with his next prototype that used a WiFi- controlled microcontroller to livestream data. Before strapping it on dogs, he tested it out on a Theo Jansen mechanism that simulated dog gait and captured linear acceleration and angular velocity to verify the accuracy of the sensors. It also allowed him to understand the right placement of the sensors. He wrote an algorithm that would eliminate noise to provide high-quality data and utilised a filter to capture just the linear acceleration, independent of gravity. The cost of the kit is about ₹5,000.
“There were other issues I had to address on the go. For instance, I got a lot of data initially which I attributed to the dog’s gait, but which was due to a wobbly sensor. Then, I had to ensure that I was capturing data only when the healthy and unhealthy dogs were moving at the same pace, which would allow for a proper comparison to be made with respect to their gaits," he says.
Once satisfied with the testing, he started collecting data over several months from different breeds of healthy dogs as well as those that suffered from a condition. He then trained Long Short Term Memory neural networks to recognise patterns in the gait and detect anomalies. Along the way, he had to constantly upgrade his skills, everything from understanding dog anatomy to experimenting with different neural networks.
“It’s an interdisciplinary project because the sensor is all physics, the movement of the dog is all biology, the artificial intelligence is all maths and then there’s the electronics involved. So getting all of these components to work together was quite a challenge, a lot of trial and error at the beginning that was quite frustrating," he says.
Bhatt made a couple of visits to the Biomechanics, Orthopaedics and Musculoskeletal Engineering (BiOME) lab at the Indian Institute of Technology Bombay to discuss his creation further.
“PawPath is a great example of how wearable sensors can be translated into real-world veterinary diagnostics. Its current application in characterising canine gait showcases its on-field deployability and is a quintessential example of potential scalable technology, which could marry concepts in biomechanics and artificial intelligence with a strong clinical relevance," says Darshan S Shah, assistant professor and head of BiOME.
The eureka moment arrived when Bhatt tested PawPath on a dog, where the output suggested that it had a high probability of arthritis. He reached out to the vet who confirmed his finding.
Bhatt has been able to achieve high accuracy in disease classification across various breeds. The results alongside the award is validation for his efforts.
In the time ahead, he wants to refine his product further to improve its utility. A smaller device will allow it to be used on smaller breeds. A longer battery life will give owners the opportunity to continuously monitor their pets even while they are away to understand behavioural changes or receive alerts in case of falls or injuries. Besides, it can also be used to track injuries in wild animals based on their movement patterns.
“I want to eventually use AI models to look at data from this device, analyse it and then hand out a simple explanation in English or Hindi on what the problem is. This can be a valuable tool for vets to make better decisions," he says.
topics
