Perfusion Parameter Map Generation from 3 Phases of Computed Tomography Perfusion in Stroke Using Generative...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Neurology

Background:

• Computed tomography perfusion (CTP) is vital for acute ischemic stroke (AIS) management but faces limitations due to complex protocols and high radiation doses.
• Reducing radiation exposure by lowering temporal sampling rates in CTP can compromise accuracy by missing critical enhancement phases.

Purpose of the Study:

• To investigate the feasibility of using a generative adversarial network (GAN) to generate accurate brain perfusion maps from a reduced number of CTP phases (mCTP).
• To evaluate the diagnostic quality and accuracy of GAN-derived perfusion maps compared to traditional CTP.

Main Methods:

• A GAN model was trained to generate perfusion parameter maps (cerebral blood flow, time-to-maximum) from three selected phases of CTP, mimicking multiphase CT angiography protocols.
• The model's performance was quantitatively assessed using structural similarity index measure (SSIM), normalized root mean squared error (nRMSE), and learned perceptual image patch similarity (LPIPS).
• Volume agreement for infarct and hypoperfusion areas was evaluated using intraclass correlation coefficient (ICC) and Spearman correlation on external datasets. Qualitative diagnostic quality was also assessed.

Main Results:

• The GAN model demonstrated high visual overlap and performance in generating cerebral blood flow and time-to-maximum maps, with high SSIM and LPIPS, and low nRMSE.
• Volume agreement for infarct and hypoperfusion areas showed good to excellent correlation (ICC and Spearman) on external datasets.
• Qualitative assessments confirmed that mCTP-derived maps were diagnostically comparable to traditional CTP.

Conclusions:

• A GAN-based approach effectively generates reliable brain perfusion maps from multiphase CTP (mCTP).
• This GAN-derived mCTP offers a promising, potentially lower-radiation alternative for diagnosing acute ischemic stroke, overcoming limitations of traditional CTP.