Overview / Usage

Detecting caries(cavities) from an X-Ray image of a patient is a tedious task and there are chances when even an experienced dentist may miss out on some small cavities. This project is targeted to automate the process cavities detection by exploring deep learning approaches for detection and segmentation. With proper optimization, the goal is to make the model scalable and Edge-ready such that it can be used by dentists at their clinics by just plugging in an Intel Neural Compute Chip to a desktop or an IoT device. This will effectively make dentists more efficient such that they can handle more patients in a given time is cavities' detection is automated and trustworthy, also it will help the patients with more accurate and fast diagnostics.

Methodology / Approach

Current standard techniques are based out of wavelets or Artificial neural networks owing to less availability of labelled data. But with proper data augmentation and Adversarial Data Augmentation, I am training a CNN based model to detect and segment cavities from the X-ray image. I am planning to use Intel Neural Compute Chip for this project to make this Edge-ready which can eliminate cloud-intervention when in use so that it protects patients' privacy and health data. I am using Tensorflow optimization for Intel to make its execution fast and real-time on a regular computer which doesn't have a high-end CPU or GPU.

Technologies Used

Currently, I am using a python environment with Tensorflow framework executed through Keras library to train and test my Neural Networks. I plan to move towards Tensorflow optimization for Intel to train and develop over Intel Neural Compute Chip.