Deep learning-based non-contrast CT imaging markers enhance post-transfer DWI core volume prediction.

Area of Science:

• Neurology
• Radiology
• Artificial Intelligence in Medicine

Background:

• Deep learning (DL) models can extract ischemic lesion imaging markers from noncontrast CT (DLNCCT).
• Predictive value of DLNCCT markers for post-transfer core volume in acute ischemic stroke remains uncertain.

Purpose of the Study:

• To evaluate if DLNCCT markers predict post-transfer core volume.
• To compare DLNCCT marker prediction with ASPECTS and CT Perfusion (CTP).

Main Methods:

• Post-hoc analysis of patients transferred for endovascular treatment (EVT).
• Quantified total lesion volume, modified net water uptake (mNWU), and severely hypodense volume using DLNCCT.
• Compared DLNCCT markers with ASPECTS and CTP for association and prediction of post-transfer core volume.

Main Results:

• DLNCCT markers (total and severely hypodense volumes) were independent predictors of post-transfer core volume.
• DLNCCT markers showed similar predictive performance to ASPECTS.
• Adding DLNCCT markers to CTP significantly improved post-transfer core volume prediction.

Conclusions:

• DLNCCT-derived total and severely hypodense volumes are valuable predictors of post-transfer core volume.
• DLNCCT markers enhance the accuracy of CTP-based prediction models for acute ischemic stroke.