Home

Welcome to the ORLIAn-Source-Code wiki!

This Wiki is about our project ORLIAn, and its model submission for the ORL/IA challenge that we decided to name El-nAI-ny (after Egyptian Arsenal player, Mohammed Elneny)

Getting started

The challenge lead, Pr. Cecile Badoual has delivered an interview on the problem that this challenge faces. This helps to provide a bit of context on what is going on here.

Keep in mind this is in French.

Challenge Details

The details of the challenge can be found on JOGL

Among ENT (Ear-Nose-Throat) oropharyngeal cancer are evolving at a high speed, especially in Northern countries, and affect predominantly young male patients with no or not excessive history of smoking or drinking. Also, an increase of ENT cancers due to Papillomavirus (HPV) infection is observed. The HPV is also found in the majority of cervix cancer and vaccination campaigns against it are now conducted. The oropharyngeal HPV+ cancers have a better prognosis certainly due to the action of the micro-environment. The micro-environment corresponds to all immune cells around the tumor cells. There are different techniques to find the presence of HPV in the cancer including in situ hybridisation. Because little is known about the behavior of these cancer, researchers and pathologists struggle to determine severity criteria and prognostic factors.

Challenge objectives

Based on pathology (tumor section/slides) data from HPV-induced orophrayngeal cancers, the ORL/IA Challenge has 2 objectives:

1. On HE (hematein/eosin) and multiparametric slides, to compare tumoral micro-environment, according to HPV quantity and tissue parameters (tumor/stromal areas).
2. To understand the role of the tissue micro-environment, with inflammatory cell activity fighting (or not) against tumor cells, and the potential importance of their spatial distribution.

The final goal is from tabular and imagery data produced, to predict the role of the micro-environment in the HPV+ oropharyngeal cancer, in tabular and imagery data produced , depending on the importance of the quantification of HPV in situ hybridisation staining.

Description of the data we are working with

This medium post does a tidy job explaining the data and how it works

Gaining access to the data

To get access to the data in Dataiku - do this:

1. Agree to the terms and conditions
2. Once agreed, wait a bit for the team to manually create an account for you. Once done, go to http://datas3.epidemium.org/ to view the data

Trouble shooting access

1. Sometimes, it takes a bit of time for the team to get to making your account. In that event, just let the team know on the #datas_and_dataiku channel

Roadmap of ORLI/A

The project will have the following stages:

1. EDA and Literature Review.
2. Baseline model created and hosted onto a private website for internal feedback from stakeholders
3. Iterations in terms of performance, keeping in mind the main objectives for this challenge
4. Deploy publicly to help aid the medical community in predicting outcomes for their patients