Within dentistry, the most important aspect of patients is their smile. As a dentist, it is our responsibility to ensure that repaired or prosthetic teeth look as realistic and esthetically pleasing as possible; color selection is a vital proponent to this process. The current methods of color and shade selection are either manual (which is tedious and inaccurate) or with equipment (which is often costly). Our inspiration was to provide the accuracy of color detection software while keeping it affordable and accessible to all of our potential users. The product born from this concept was Hū, a shade-guide application for mobile devices, which can be used by both dentists and dental students alike.
What it does
Currently, the android application can take a picture of a patient's teeth using a conical attachment that minimizes light distortions. Manually, the user would take the image and crop a circular shape that runs through a visual basic script that averages the HSV color of each pixel of the tooth. It then uses a pre-calculated database of teeth shadings to determine the closest shading that the filling could use.
How I built it
We used Android Studio to create the android application and Visual Studio to create the color calculation script. We also used Python to automatically perform image processing to determine the teeth's edges and surfaces. There were difficulties in automatically determining closed surfaces to average the pixel colors so the user must manually crop a circular area for the program to process.
Challenges I ran into
We had difficulties in running Python code on the Android app using Kivy. We also had problems in writing algorithms to run automatic detection in the Python code. We had a 3D model for the conical mold, but ran into many issues with the 3D printers so we had to create a makeshift paper model.
What's next for Hu
Ideally, Hu would be able to take an image and using automatic image processing and some minor machine learning, be able to determine individual teeth and match the colors to shadings without any further user interaction. We would also want to make a plastic 3D printed mold for the conical attachment because it is significantly more stable than the makeshift paper version.