A PET-Guided Framework Supports a Multiple Arterial Input Functions Approach in DSC-MRI in Acute Stroke

Arterial input functions in dynamic susceptibility contrast MRI (DSC-MRI) in longitudinal evaluation of brain metastases

BACKGROUND

Dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) could be helpful to separate true disease progression from pseudo-progression in brain metastases when assessing the need for retreatment. However, the selection of arterial input functions (AIFs) is not standardized for analysis, limiting its use for this application.

PURPOSE

To compare population-based AIFs, AIFs specific to each patient, and AIFs specific to every visit in the longitudinal follow-up of brain metastases.

MATERIAL AND METHODS

Longitudinal data were collected from eight patients before treatment (6 of 8 patients) and after treatment (6-17 visits). Imaging was performed using a 1.5-T MRI system. Lesions were segmented by subtracting precontrast images from postcontrast images. Cerebral blood volume (rCBV) and cerebral blood flow (rCBF) were computed, and Pearson's product moment correlation coefficients were calculated to evaluate similarity of DSC parameters dependent on various AIF choices across time. AIF shape characteristics were compared. Parameter differences between white matter (WM) and gray matter (GM) were obtained to determine which AIF choice maximizes tissue differentiation.

RESULTS

Although DSC parameters follow similar patterns in time, the various AIF selections cause large parameter variations with relative standard deviations of up to ±60%. AIFs sampled in one patient across sessions more similar in shape than AIFs sampled across patients. Estimates of rCBV based on scan-specific AIFs differentiated better between perfusion in WM and GM than patient-specific or population-based AIFs (P ≤ 0.02).

CONCLUSION

Results indicate that scan-specific AIFs are the best choice for DSC-MRI parameter estimations in the longitudinal follow-up of brain metastases.

Image-to-image generative adversarial networks for synthesizing perfusion parameter maps from DSC-MR images in cerebrovascular disease

Stroke is a major cause of death or disability. As imaging-based patient stratification improves acute stroke therapy, dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) is of major interest in image brain perfusion. However, expert-level perfusion maps require a manual or semi-manual post-processing by a medical expert making the procedure time-consuming and less-standardized. Modern machine learning methods such as generative adversarial networks (GANs) have the potential to automate the perfusion map generation on an expert level without manual validation. We propose a modified pix2pix GAN with a temporal component (temp-pix2pix-GAN) that generates perfusion maps in an end-to-end fashion. We train our model on perfusion maps infused with expert knowledge to encode it into the GANs. The performance was trained and evaluated using the structural similarity index measure (SSIM) on two datasets including patients with acute stroke and the steno-occlusive disease. Our temp-pix2pix architecture showed high performance on the acute stroke dataset for all perfusion maps (mean SSIM 0.92-0.99) and good performance on data including patients with the steno-occlusive disease (mean SSIM 0.84-0.99). While clinical validation is still necessary for future studies, our results mark an important step toward automated expert-level perfusion maps and thus fast patient stratification.

Simulated perfusion MRI data to boost training of convolutional neural networks for lesion fate prediction in acute stroke

The problem of final tissue outcome prediction of acute ischemic stroke is assessed from physically realistic simulated perfusion magnetic resonance images. Different types of simulations with a focus on the arterial input function are discussed. These simulated perfusion magnetic resonance images are fed to convolutional neural network to predict real patients. Performances close to the state-of-the-art performances are obtained with a patient specific approach. This approach consists in training a model only from simulated images tuned to the arterial input function of a tested real patient. This demonstrates the added value of physically realistic simulated images to predict the final infarct from perfusion.