aiDent

aiDent is a personalized orthodontist which prevents misdiagnosis in the dental industry by leveraging the use of advanced machine learning models for detecting various diseases using X-Ray img

Comment

Inspiration

The foundations for building an application that detects misdiagnosis in the orthodontal industry came when one of our teammates (who had a prior experience in the industry) noticed that dental industry although being huge, was heavily saturated which kept the door shut for the use of modern tools (AI and image detection). During modern times, every industry across the world is benefitting by the use of AI. The dental industry being the biggest healthcare industries suffers heavily due to the lack of new models and resources which in turn affects the insurance providers, patients and the dentist. According to National Library of Medicine, general dentist misdiagnosed 45.9% of the patients during research. Additionally, according to ADA Member Advantage, a person from the age segment of 20-70 spends minimum $51,000 in their life; if 45.9% of the detected diseases are misdiagnosed an average person spends $23409 dollars in their lifetime in waste. This expense is indirectly paid by the insurance companies for the treatment, which creates a huge waste of patient’s time resources and insurance’s along with customer’s billions of dollars in vain. This number being huge heavily motivated us to develop a solution which could lower the misdiagnosis caused as these misdiagnoses induces a mental instability in patients. Getting into a greater depth regarding misdiagnosis helped us realize that root canal along with some other diseases was one of the most operated procedures in the world, with about 15 million people getting the treatment every year in the United States. Viewing the gravity of the situation, deeply inspired us to break this stigma and give the patients as well as the dentist a good experience while their dental treatments.

What does it do?

aiDent is a personalized orthodontist which prevents misdiagnosis in the dental industry by leveraging the use of advanced machine learning models for detecting various diseases using X-Ray images. Our web app is user and business focused application which uses a simple easy to use API and breaks down the image into its individual diseases by individual tooth. Running advanced machine learning algorithms in the backend helps us ensure that the product is scalable over time. When a customer visits a dental clinic and their orthodontist defect’s a disease during their appointment; they click and X-RAY to ensure that the teeth is infected usually by themselves or by private players. After receiving an X-RAY, the user can visit our website and make an account. The online X-RAY images could be later be uploaded into the model and the backend machine learning algorithm and extract an image that can detect specific diseases into specific teeth on the X-RAY. If there is a misdiagnose present, our model will help the patient recognize that there is a disease present on a specific tooth in the X-RAY. After knowing about the exact cause, they can communicate effectively to their orthodontist and figure out their exact cause. This helps the insurance agency’s as they do not have to spend money on misdiagnose treatment which enables them to save their money and resources effectively. Furthermore, the patents do not have to go through the pain and the mental trauma of going through a misdiagnose surgery.

Tools involved in building it

We used YoloV8 and YoloV6 pretrained models to train our machine learning models in google collab and using basic open cv framework in Jupiter notebook. In additionally, for frontend we used React.js framework for designing and TailWindCSS for integration. Backend we used Flask and TailWindCSS.

Accomplishments that we're proud of

aiDent has the power and potential to revolutionize the dental industry owing to it’s clean and scalable machine learning algorithms and easy to use friendly UI for the user. Thousands of people all across United States and numerous people around the world struggle because of misdiagnosis every single day. Misdiagnosis not only creates a physical threat on a both but also create a negative impact on the patient’s mind regarding treatments due to their prior experiences. If a patient gets treated on a misdiagnosis, the amount of mental trauma on the patient after the medical procedure is unbelievable. Our product can solve this issue completely and give the patient the adequate treatment that they deserve and prevent the mental trauma altogether.’ Our product will also prove to be a huge boom for the insurance industries as they could use it as an verification model. Additionally, if we collaborate with the insurance companies; we would be able to collect additional data set and further increase the precision and accuracy of our models. This could prevent important resources like money and man power to significantly reduce and ensure that people get the right treatment for their disease. In addition, creating this idea within 24 hours in a hackathon was a dream come true for us. After working continuously throughout the entire duration of the hackathon nonstop and not regretting a part of it portrays our pride and passion towards building it. Our mentors really helped us integrate all of our model along the way and we really appreciate the help and valuable advice that they gave us which resulted into such a cutting-edge software with a strong UI foundation.

Challenges we ran into

A major challenge that we ran into was collecting data for training our Machine Learning models and integrating our machine learning algorithms, backend architecture and frontend interface. While searching for data on the internet, we faced a challenge of collecting reliable and well formatted data as our classification machine learning model dependent highly on the data used. Additionally, due to privacy reasons medical colleges and clinics could not share us the datasets to us and were very supportive about the mission although did not have the permission to share the data outside their organizations.

Secondly, after creating and heavily testing our machine learning models, back-end framework and fronted interface; we struggled to integrate them together in order to create a complete web app. After talking and gaining valuable advice from our mentors, we finally figured out an API’s which enabled us to integrate all the individual parts together into one tight knit environment.

Thirdly, we have a couple of cases where our machine learning model is unable to classify the images; though very less amount of time, it is caused due to limited data and lack of advanced filtering system (which is very difficult to implement in 24 hours). We can fix this using advanced classification on our data set, getting additional data set and further enhance the accuracy of the model.

Future Plans

After attending this hackathon, every one of us has understood of launching and deploying this software in the industry as it could benefit millions of people around the world and could help them prevent misdiagnosis which creates a metal instability, severe trauma and phobia of getting treatment caused due to their negative experience; on the other hand it could help dentists to ensure that their patients are getting the right treatment, avoid potential lawsuits and give their end customers (patients) a very fluid experience throughout their treatment. This creates a win-win situation for both the doctors and their patients. We strongly think that this has a wide application in the insurance space as they are paying billions of dollars every year for their clients respective treatments. If I algorithm ideally works, the insurance companies can save up to 45.9% of its total money and resources spent which can furthermore help them control their cost and ensure that their clients are getting the right treatment necessary. Additionally, we could provide insurance companies with a verification modular system which requires to run the patients X-Rays using our application. This will not only help the insurance companies save their valuable finance but will also help us collaborate with the companies to extract additional data set from them to train and increase accuracy of our model. In future, we would like to collaborate with a dental organization which could provide us additional and reliable data set helping us scale our product and increasing the reliability and usability. Additionally, we would be able to further collect additionally data set from our users which we could use to further enhance our neural nets and create a scalable product. We strongly think that people all across the world are in need for our product and we strongly want to help them prevent misdiagnosis and leave millions of people with a smiling face!

Built With

Share this project:

Updates