Radiology: Diagnostic value of synthetic FLAIR sequences in acute ischemic stroke

It is known that the time to symptom onset is unknown in 27% of acute ischemic stroke (AIS) patients
.

However, the assessment of stroke onset time is one of the important references for the implementation of current reperfusion therapy .

It is known that the time to symptom onset is unknown in 27% of acute ischemic stroke (AIS) patients

Because time is of the utmost importance in the treatment of AIS, the development of a short scan protocol is crucial for further treatment

Based on the assumption that key T2-weighted signal changes are embedded in weights with b0 and b values ​​of 1000 sec/mm ² 2 (hereafter b1000), we hypothesize that synthetic FLAIR images generated from DWI alone by using deep learning methods can be compared with FLAIR The collected images (hereinafter referred to as real FLAIR) compete for DWI-FLAIR mismatch judgment

A study published in the journal Radiology compared deep learning-derived synthetic FLAIR and real FLAIR images in displaying and identifying DWI-FLAIR mismatches in AIS patients imaged within 4.
5 hours of stroke onset , for reducing The scanning time of acute stroke patients provides the possibility .

This retrospective study included all pre-treatment and early follow-up (<48 hours after symptom onset) MRI images including DWI (b=0-1000 ) of AIS patients undergoing reperfusion therapy between January 2002 and May 2019 sec/mm2) and FLAIR sequences .
On the training set (80%), a generative adversarial network is trained to produce synthetic FLAIR with DWI as input .
In the test set (20%) , synthetic FLAIR is computed without real FLAIR .
Four independent readers evaluated the DWI-FLAIR mismatch for these two FLAIR datasets .
The interobserver statistical repeatability and agreement of the DWI -FLAIR mismatch between synthetic FLAIR and true FLAIR were assessed using kappa statistics .
In patients with known time to onset, the McNemar test was used to compare the sensitivity and specificity of identifying AIS within 4.
5 hours .

The study included 1416 MRI examination images (861 patients; median age, 71 years [interquartile range, 57-81 years]; 375 men), yielding 1134 and 282 scans for training and testing , respectively set .

Regarding the DWI-FLAIR mismatch, the interobserver reproducibility of real and synthetic FLAIR was high (κ = 0.
80 [95% CI: 0.
74, 0.
87] vs0.
80 [95% CI: 0.
74, 0.

The study included 1416 MRI examination images (861 patients; median age, 71 years [interquartile range, 57-81 years]; 375 men), yielding 1134 and 282 scans for training and testing , respectively set .
Regarding the DWI-FLAIR mismatch, the interobserver reproducibility of real and synthetic FLAIR was high (κ = 0.
80 [95% CI: 0.
74, 0.
87] vs  0.

 

Comparison of DWI-FLAIR mismatch by using acquired FLAIR sequences (true FLAIR) and synthetic FLAIR in patients with acute ischemic stroke .

(A) A 46-year-old female with sudden right hemiparesis, aphasia, and dysarthria with mismatched DWI-FLAIR images on both synthetic FLAIR and real FLAIR images .
On DWI scans obtained 2 hours after onset, areas of hyperintensity appeared in the left middle cerebral artery supply area , with no corresponding signal changes on real or synthetic FLAIR images .
(B) 47-year-old man with sudden right hemiparesis and no DWI-FLAIR mismatch on synthetic FLAIR and real FLAIR images .
On a DWI scan obtained 5 hours after onset, a small hyperintensity (arrow) was seen in the left internal capsule , also visible in real and synthetic FLAIR images .

Comparison of DWI-FLAIR mismatch by using acquired FLAIR sequences (true FLAIR) and synthetic FLAIR in patients with acute ischemic stroke .
(A) A 46-year-old female with sudden right hemiparesis, aphasia, and dysarthria with mismatched DWI-FLAIR images on both synthetic FLAIR and real FLAIR images .
On DWI scans obtained 2 hours after onset, areas of hyperintensity appeared in the left middle cerebral artery supply area , with no corresponding signal changes on real or synthetic FLAIR images .
(B) 47-year-old man with sudden right hemiparesis and no DWI-FLAIR mismatch on synthetic FLAIR and real FLAIR images .
On a DWI scan obtained 5 hours after onset, a small hyperintensity (arrow) was seen in the left internal capsule , also visible in real and synthetic FLAIR images .
(C) An 86-year-old female with sudden right hemiparesis at 5 hours and 10 minutes before MRI with no DWI-FLAIR mismatch on synthetic FLAIR and real FLAIR images .
On DWI scans obtained 5 hours and 10 minutes after onset, a large area of ​​hyperintensity appeared in the area supplied by the left middle cerebral artery , also visible on real and synthetic FLAIR images .
Notably, white matter hyperplasia (arrow) was equally visible on real and synthetic FLAIR images (Fazekas score of 2)

This study demonstrates that in patients with acute ischemic stroke , synthetic FLAIR sequences can achieve diagnostic performance similar to real FLAIR sequences for DWI- FLAIR mismatch assessment, and thus can replace FLAIR sequences for clinically relevant stroke MRI scans Program time is shortened .

Original source :

Joseph Benzakoun , Marc-Antoine Deslys , Laurence Legrand , et al .

Synthetic FLAIR as a Substitute for FLAIR Sequence in Acute Ischemic Stroke .
DOI: 10.
1148/radiol.
211394

Joseph Benzakoun Marc-Antoine Deslys Laurence Legrand ,et al Synthetic FLAIR as a Substitute for FLAIR Sequence in Acute Ischemic Stroke 10.
1148/radiol.
211394 10.
1148/radiol.
211394Leave a comment here