Ischemic Stroke Lesion Segmentation Challenge - ISLES'24

Final, post-treatment infarct segmentation from pre-treatment acute imaging and clinical data

---------------------------------------------------------------------------------------------------------------------------------------------

NEWS:
[01.04]
- Stay tuned! We are working on this website and we'll gradually include more info about this ISLES edition!  🤓🧠
- Our challenge was accepted at MICCAI'24! 
---------------------------------------------------------------------------------------------------------------------------------------------

BACKGROUND

Clinical decisions regarding the treatment of ischemic stroke patients depend on the accurate estimation of core (irreversibly damaged tissue) and penumbra (salvageable tissue) volumes (Albers et al. 2018). The clinical standard method for estimating perfusion volumes is deconvolution analysis, consisting of i) estimating perfusion maps through perfusion CT (CTP) deconvolution and ii) thresholding the perfusion maps (Lin et al. 2016). However, the different deconvolution algorithms, their technical implementations, and the variable thresholds used in software packages significantly impact the estimated lesions (Fahmi et al. 2012). Moreover, core tissue tends to expand over time due to irreversible damage of penumbral tissue, with infarct growth rates being patient-specific and dependent on diverse factors such as thrombus location and collateral circulation. Understanding the core's growth rate is clinically crucial for assessing the relevance of transferring a patient to a comprehensive stroke center based on transport times (Robben et al. 2020). Moreover, since not every reperfusion treatment with mechanical thrombectomy achieves complete reperfusion, predicting infarct growth might provide interventional radiologists with insights into the potential benefits of additional reperfusion attempts. Therefore, anticipating the temporal core evolution from acute imaging data is key for clinical decision-making. 

TASK

This challenge aims to segment the final stroke infarct from pre-interventional acute stroke data. 

Inputs:

• Acute CT images (NCCT, CTP and CTA)
• Tabular data (demographic and clinical data).

Outputs:

• Binary infarct segmentation mask. 

Schedule

• Release of Training data (1st batch): 15th of May 2024

• Release of Training data (2nd batch): 15th of June 2024

• Opening of submission system for dockers: 1st of July 2024

• Closing of submission system for dockers: 15th of August 2024

Registration

(!) Verified accounts are mandatory to be accepted to the challenge.