She won the Google contest. Her solution just a photograph can identity Down Syndrome.

A teenager thought about something that slips the minds of even the educated big heads. Something that's often ignored and something that's not stressed upon. Everyone knows the pain behind it but not anyone pays any heed. "Down Syndrome", a genetic disorder that causes developmental and intellectual delays. Thanks to Google's approaches to bring out the best in common people who can innovatively think for the betterment of the society using technological eye, a 15-year-old girl developed a machine learning project that identifies children with down Syndrome at an early stage. She is Saanvi Mehra, a class X student from Noida and this is the story of her landmark innovation. 

Fifteen-year-old Saanvi Mehra is from Noida and goes to The Shriram Millennium School, Noida. This class X girl participated in Google India Code to Learn contest 2020 and emerged victorious. Her project is helpful to identify children with Down Syndrome through their photographs. 

Early the Detection, Quick the Treatment: 

According to Saanvi's statements in her blog, a large number of children globally, as well as in India are born with Down Syndrome. In India, roughly one out of every thousand children, that is, over 30,000 children are being born every year with Down Syndrome. The survival rate of children with this syndrome in India is as less as 44%. Lack of medical facilities that lead to non-diagnosis or late-diagnosis and the expensive medical procedures to detect this syndrome are the main causes for high mortality rate in India. A highly expensive ultrasound scan that can detect children with Down Syndrome well ahead of their birth is beyond the reach of common people. This chromosomal abnormality lands children on the risk of congenital heart defect, developing pulmonary hypertension, and other medical conditions related to hearing and vision. While many of these are treatable, one should detect these conditions early. "All infants with Down syndrome need to be evaluated by a pediatric cardiologist. Regular evaluations by an otolaryngologist (ear, nose, and throat doctor), audiologist, and an ophthalmologist are necessary to detect and correct any problems before they affect language and learning skills.", mentions Saanvi in her blog. 

Here is where Saanvi took charge and worked upon a solution for this.

Just a Click Away: 

Saanvi designed a project based on a machine learning model. What if you can just click a picture of you’re newly born and with that, you can detect if he suffers from Down Syndrome without the need of any expensive medical procedures? With a smartphone in everyone's hand, every parent can very well click a photograph of their child which can later be dropped in a free tool that looks for the signs of Down Syndrome in the child. Early detection not only reduces the death of the child but also aids in timely treatment gifting them new lives. 

Saanvi's Working Idea: 

What can a photograph do in this case? Well, it's not only the photograph. Saanvi used Machine learning techniques that facilitate the recognition of facial dysmorphic features associated with genetic syndromes. She used Cloud AutoML Vision Multi Label classification model to assign labels and train the model. She mentions that, once her model was deployed, she started with almost 100 images die the two labels, tested and validated with multiple images. Saanvi went on to get relevant data from medical forms and authorities in order to increase the accuracy of her model making it suitable for multiple regions. Studies show that by understanding the facial features associated with children who fell prey to this disorder, one can detect children with Down Syndrome. "Some common characteristics present in people with the disorder include distinctive facial features, such as slanting eyes, small chin, round face, flat nasal bridge, Brushfield spots in the iris, abnormal outer ears, and a flattened nose." She says. Saanvi designed her model and to ensure that it is not region-specific, her dataset consisted of images of children from across the world including India. Amazingly, this model has given accurate results in detecting DS children without missing on at least one. Saanvi's innovation saves close to 10k lives in India if properly taken up. 

The contest organisers, Google, went into a state of amazement after looking at the girl's project and declared her the winner in the Class IX-X category. 

We are proud of Saanvi and her noble project which will save thousands of children who fall prey to Down Syndrome. She has taken forward the values of our institution and we are ready to support her in any way we can, said Ms Uttara Singh, Director, The Shriram Millennium schools. The director also said that other students shall also be encouraged towards AI and programming. 

  CheckOut Her Blog