Cepheo is an innovative DL driven online orthodontic and orthognathic platform for dental clinicians.

We developed a node.js and flask APIs connected to MongoDb database, through which we deployed the best model. It automatically performs landmark positioning. It can be used as an online diagnostic software that reduces the time and effort of orthodontists involved in the orthodontic case and diagnosis.

• The dentical clinician can add new patient via upload page.
• View the all patients data in one page in the repository part.
• Through the repository section also the dentical clinician can also update, delete the medical record for each patient.
• After the analysis is done, the patient record can be saved as a PDF report or image.
• We also provide in our platform help page that contains frequently asked questions and short instruction video in our youtube channel to provide information of everthing till getting analysis done.

Client: HTML, CSS, JS

Server: Node, MongoDB, Flask

Model: UNet, Inception, Resnet

Detection of cephalometric landmarks with Deep learning models. Dataset:

• 400 X-ray images each image contains 19 landmarks.
• Excel sheets containg X and Y positions for each landmark in each image.

• Binary Mask Trial Unet.ipynb The mask used for segmentation is a binary 1 channel array, where 1's represent landmarks location and 0's represent the background.

• Unet_and_get_ROIs_for_Stage2.ipynb The mask used for segmentation is a 20 channels mask, where each channel represent a landmark and the last channel represent the background. Each channel was filled using the following equation.

• resnet_inception_final.ipynb

• resnet_rois.ipynb

Install node.js 👉 Download

To use the Model 👉 Download

Install

• FlaskAPI: Flask, TensorFlow, OpenCv, Numpy
• NodeAPI: npm install command

Start the node server

  node app.js

Add model in models folder then start flask server

  run main.py

Start the client

  run index.html

HTML, CSS, Javascript, Node, Flask, MongoDB and deep learning

• Sample of saved analyzed image

• Sample of our patient medical record

For support, email [email protected] or join our community on facebook.

If you have any feedback, please reach out to us at [email protected]

• @tasneemkhaled
• @maryammahmoud
• @hadeerkhaled
• @nouranMuhammed
• @Lailahamdy
• @nadaAhmed
• @marwahelmy
• @hanaasalah
• @nadaabdelrahman