# Grand-Challenge Request Form

**Title :** Light my cells : Bright-field to Fluorescence Challenge

**Acronym :** lightmycells

**Contact email :** [e-mail]

**Abstract:**

France Bio-Imaging's first challenge is to develop a method for imaging mammals with targeted fluorescence at the cellular level using bright-field, phase-contrast and/or DIC microscopy without using markers. This method aims to manage the diversity of cells, instruments and markers.

**Start Date :** 15/12/2023 <span style="color:red"> ?? </span>
**End date :** 27/04/2024 <span style="color:red"> ?? </span>

[Timeline](https://hackmd.io/_uploads/BkrTDeigT.png)

**Organizers :**
Emmanuel Faure - researcher at LIRMM & FBI.data mission officer
Dorian Kauffmann - project engineer at LIRMM, working full time for the FBI Challenge project
Guillaume Gay - research engineer at LIRMM, working full time for the FBI.data project
Edouard Bertrand - Scientific Director - FBI team leader
Thomas Walter - Professor in Artificial Intelligence for the life sciences at Mines Paris and Director of the Centre for Computational Biology (CBIO)

**Afiliated event :** (hopefully) ISBI2024 <span style="color:red"> ?? </span>

**Structured challenge submission form :** No

**Task types :** <span style="color:red">Registration </span>

**Structures :** Nucleus / Mitochondria

**Modalities :** Bright-field, phase contrast and fluorescence microscopy (DIC)

**Challenge setup :**

+ phases :
+ Phase 1 preliminary Test
+ Phase 2 final Test

+ tasks :
+ Task 1 : Im to Im fluo Nuc & Mit fluo
+ Task 2 : stack Im to Best focus fluo Nuc & Mit

<span style="color:gray"> -> 2x2 leaderboards </span>

**Data set :**

+ training : <span style="color:red"> ToSee </span>
+ test : <span style="color:red"> ToSee </span> (to upload on GdChlg)

see how to benchmark database, k-fold chosen, sets struct etc.

**CC-BY licence agreement :** Yes

**Submission assessment :**
+ metrics : <span style="color:red"> ToSee </span>
SSIM, FSIM, Coefficient Dice-Jaccard, Coefficient de Pearson ?
1 for nucleus + 1 for mito(based on intensity)

Importance of nucleus interior -> check intensity ?

+ ranking : <span style="color:red"> ToSee </span>
Accuracy, Precision, Recall, + test metrics ?
+ (stats)

**Challenge publication :**

+ plan to coordinate a publication of the challenge results :
+ Paper 1 only on the Database
+ Paper 2 on the Challenge + code

**Code availability :** <span style="color:red"> Open-source license -> which : MIT, Apache 2.0, etc.</span>
Tout le monde peut part. et gagnant seront afficher par ordre peu importe type code, mais seul participants qui ont open source pourront gagner un prix (+accompgnement code accessible). -> a la soumission (pas en fct du score)

**Algorithm inputs :** Bright Field or Phase Contrast Microscopy Image of mammalian live cells. <span style="color:red"> OME-TIFF only ?? </span>

**Algorithm outputs :** Specific Fluorescence Image of associated mammalian Nucleus or Mitocondria.

**Expected number of teams :** 50
**Number of tasks :** 2
**Average size of test image in mb :** ToSee 350 mb (~42 Mo)
**Average algorithm job run time in minutes :** 5min

**Phase 1**
**Expected number of submissions per team to Phase 1 :** 5
**Phase 1 number of test images :** <span style="color:red"> 10 ?? </span>

**Phase 2**
**Expected number of submissions per team to Phase 2 :** 2
**Phase 2 number of test images :** <span style="color:red"> 100 ?? </span>

Budget for hosting challenge in euros : ~8.000 euros (estimation of 8.080euros)

**Comments : **