Inspiration

  • Aiding the doctor in planning the ideal patient care, by supporting the tedious and error-prone cancer classification with Deep Neural Networks.
  • Enabling patients all around the world - even in remote places - to receive (better) care through a new decentralized apeer-based ecosystem

What it does

The doctor can upload a tissue sample to Apeer, where it is automatically processed and where a report is generated for the doctor, informing him about the decision in cancer or no_cancer together with a confidence metric.

How we built it

Divided up the work to match compatibility with our team members' skills. The Deep Learning models were trained on the Azure cloud VMs and then a docker image was created with the generated weights, which was then put in the APEER pipeline.

Challenges we ran into

Setting up Jupyter notebooks with azure, building a docker for the apeer pipeline

Accomplishments that we're proud of

Set up a sophisticated pipeline for training out models in the Azure cloud VMs. Used dynamic resources available to train 4 models simultaneously in 4 different GPU simultaneously. Tested the robustness of the Azure services and were quite happy with it.

What we learned

We learned how to learn how to put Deep Learning models in production through the APEER platform. Most importantly learned many new technologies while trying to help out team members and building our product in such a short period of time.

What's next for DOCTAR - Deep Optical Cancer Tissue Apeer Research

Enabling rapid adoption of the ecosystem by pathologists, doctors, and patients through successively developing software services built upon the ideas of transparency and interpretability.

Built With

  • azure
  • apeer
  • tensorflow
  • keras
  • love
  • zeiss
  • vs-code-editor
Share this project:
×

Updates