Smart lung disease diagnosis and personalized follow-up
Tagline
AI² – Smart lung diagnosis and personalized follow-up
Summary
Our innovative solution aims to leverage AI (machine learning) on two streams (AIxAI=AI²), lung patient diagnostic, and hospital administrative workflow, to both streamline specialist workload and patient experience.
We are improving these processes by:
- using a prediction model based on lung X-rays, to determine a lung patient pathology
- automating coordination between diagnosis and booking assistant tools (CRM), to propose a relevant personalized follow-up appointment with a lung specialist
We also remove friction points in the process by using user-friendly app for the patient (e.g. to set reminders, localization of appointment, etc.)
Link to the one page article
Click here for the more detailed approach
Additional Resources
XRay Analysis algorithm details
Ressources: CheXpert: Irvin, Jeremy, Pranav Rajpurkar, Michael Ko, Yifan Yu, Silviana Ciurea-Ilcus, Chris Chute, Henrik Marklund, Behzad Haghgoo, Robyn Ball, und Katie Shpanskaya. „CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison“. arXiv preprint arXiv:1901.07031, 2019.
ResNet: He, Kaiming, Xiangyu Zhang, Shaoqing Ren, und Jian Sun. "Deep residual learning for image recognition". In Proceedings of the IEEE conference on computer vision and pattern recognition, 770–778, 2016.
We tried the VGG network, but it didn´t perform as well: Simonyan, Karen, und Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556, 2014.
Technical Specifications
Click here for the technical specifications
Installations & Reproduction Detailed Instructions
Click here to access installation instructions
Click here to access reproducibility instructions
Team Members
Benedikt Jordan Linkedin, Benedikt.jordan@posteo.de
Anass Bellachehab Linkedin, ans.bellache@gmail.com
Allwyn Joseph Linkedin, allwyn@azmed.co
Roman Sztergbaum Linkedin, Github, rmscastle@gmail.com
Noémie Héroin Linkedin, noemie.heroin@gmail.com