주제 1 뇌경색 후 뇌출혈 예측
Scientific overview (연구 배경 및 중요성)
Hemorrhagic transformation (HT) is one of the major complications of
acute ischemic stroke and is reported to occur in up to 37.5% of patients following
natural or mechanical recanalization of the occluded cerebral arteries.
It deters the use of antithrombotic treatment and is recognized as one of the poor
prognostic markers. Prediction of HT guides establishing treatment strategies and enables
patient stratification and prediction of prognosis.
Clinical parameters as well as imaging parameters have been used to predict HT in acute
ischemic stroke. Developing artificial intelligence-based models using initial MR images to
predict HT may guide establishing treatment strategies and enable patient stratification and
prediction of prognosis.
Challenge questions (문제 정의)
Develop an AI model to predict hemorrhagic transformation after cerebral infarction using the initial image information of acute stroke patients.
Data description (데이터 설명 – 데이터 셋 구성, 형식, 특징)
Data collection from a single-center (Asan Medical Center) will be provided as the training and validation data for this year’s HeLP challenge - Stroke Hemorrhagic Transformation Prediction.
- (1) Data Set : 200 of acute stroke patients (120 and 80 for Training and Validation, respectively)
- (2) Clinical Data and Label : Occurrence of hemorrhagic transformation (training data only)
- (3) Image Data : Diffusion-weighted MRI (DWI), Perfusion-weighted image (PWI), fluid attenuated inversion recovery (FLAIR) and T2-MRI (gradient-echo MRI or T2*-weighted MRI)
MRI scans were performed using 1.5-T clinical whole body scanner (Avanto, Siemens) with a standard head coil. For each patient, an acute stroke MRI protocol including DWI, PWI, FLAIR, GRE (or T2*-weighted image) sequences was used. ADC map and PWI maps (K2, MTT, rBF, rBV, TMAX, tMIP and TTP) were automatically created from DWI and PWI scans using the built-in software (Olea Medical).
Not provided.
The brain MRI images and infarct lesion mask are provided in compressed NifTi format (.nii.gz). For each patient, the provided FLAIR, PWI, and T2-MR (gradient-echo MRI or T2*-weighted MRI) images was co-registered to its native space (i.e., DWI acquisition space). Registration was applied using ANTs Toolbox: from its native space to Montreal Neurological Institute (MNI) space with a resolution of 2 mm × 2 mm × 2 mm (91 x 109 x 91 voxels in x-, y-, and z-axis).
Each patient have two folders based on preprocessing (i.e., DWI space or MNI space). Each folder contains 14 files of images as followed :
[DWI_Space] Folder (14 files): Brain mask was estimated in DWI (dwi-b0 image). FLAIR, PWI, T2-MR images were co-registered to DWI space (i.e., same resolution and dimension space in each patient).- (1) dwi_adc_brain.nii.gz
- (2) dwi_b0_brain.nii.gz
- (3) dwi_b0_brain_mask.nii.gz
- (4) dwi_b1000_brain.nii.gz
- (5) flair_in_b0_brain.nii.gz
- (6) pwi_K2_in_b0_brain.nii.gz
- (7) pwi_MTT_in_b0_brain.nii.gz
- (8) pwi_rBF_in_b0_brain.nii.gz
- (9) pwi_rBV_in_b0_brain.nii.gz
- (10) pwi_ref_in_b0_brain.nii.gz
- (11) pwi_TMAX_in_b0_brain.nii.gz
- (12) pwi_tMIP_in_b0_brain.nii.gz
- (13) pwi_TTP_in_b0_brain.nii.gz
- (14) t2star_in_b0_brain.nii.gz (or gre_in_b0_brain.nii.gz)
- (1) dwi_adc_in_MNI_brain.nii.gz
- (2) dwi_b0_brain_mask_in_MNI.nii.gz
- (3) dwi_b0_in_MNI_brain.nii.gz
- (4) dwi_b1000_in_MNI_brain.nii.gz
- (5) flair_in_MNI_brain.nii.gz
- (6) pwi_K2_in_MNI_brain.nii.gz
- (7) pwi_MTT_in_MNI_brain.nii.gz
- (8) pwi_rBF_in_MNI_brain.nii.gz
- (9) pwi_rBV_in_MNI_brain.nii.gz
- (10) pwi_ref_in_MNI_brain.nii.gz
- (11) pwi_TMAX_in_MNI_brain.nii.gz
- (12) pwi_tMIP_in_MNI_brain.nii.gz
- (13) pwi_TTP_in_MNI_brain.nii.gz
- (14) t2star_in_MNI_brain.nii.gz (or gre_in_MNI_brain.nii.gz)
* MTT = mean transit time, rBF = relative blood flow, rBV = relative blood volume, TMAX = time-to-maximum, tMIP = time maximum intensity projection, TTP = time-to-peak
Please refer an article to understand the meaning of PWIs (Demeestere et al., Review of Perfusion Imaging in Acute Ischemic Stroke: From Time to Tissue, Stroke 2020:51(3), https://doi.org/10.1161/STROKEAHA.119.028337.
Evaluation matrix (정량적 평가 방법)
(1) Requirements for participants in submission (excel)
- Submit the estimated probability of hemorrhagic transformation by the developed methods
(2) Evaluation
- Area-under-the-curve (AUC) using the receiver-operating-characteristic (ROC) between submitted probability and labeled occurrence of hemorrhagic transformation in validation data
