MY CONTENT

Convolutional neural networks were trained on combinations of DWI, ADC, and low b-value-weighted images from 116 subjects. The performances of the networks (measured by the Dice score, sensitivity, and precision) were compared with one another and with ensembles of 5 networks. An ensemble of convolutional neural networks trained on DWI, ADC, and low b-value-weighted images produced the most accurate acute infarct segmentation over individual networks. Automated volumes correlated with manually measured volumes for the independent cohort.

The purpose of this study was to introduce a multiphase MRA collateral map derived from time-resolved dynamic contrast-enhanced MRA and to verify the value of the multiphase MRA collateral map in acute ischemic stroke by comparing it with the multiphase collateral imaging method (MRP collateral map) derived from dynamic susceptibility contrast-enhanced MR perfusion. The authors generated collateral maps using dynamic signals from dynamic contrast-enhanced MRA and DSC-MRP in 67 patients using a Matlab-based in-house program and graded the collateral scores of the multiphase MRA collateral map and the MRP collateral map independently. Interobserver reliabilities and intermethod agreement between both collateral maps for collateral grading were tested. The interobserver reliabilities forcollateral grading using multiphase MRA or MRP collateral maps were excellent. They conclude that the dynamic signals of dynamic contrast-enhanced MRA can generate multiphasecollateral images and show the possibility of the multiphase MRA collateral map as a useful collateral imaging method in acute ischemic stroke.