Decomposing cerebral blood flow MRI into functional and structural components: A non-local approach based on prediction

Abstract

We present RIPMMARC (Rotation Invariant Patch-based Multi-Modality Analysis aRChitecture), a flexible and widely applicable method for extracting information unique to a given modality from a multi-modal data set. We use RIPMMARC to improve interpretation of arterial spin labeling (ASL) perfusion images by removing the component of perfusion that is predicted by the underlying anatomy. Using patch-based, rotation invariant descriptors derived from the anatomical image, we learn a predictive relationship between local neuroanatomical structure and the corresponding perfusion image. This relation allows us to produce an image of perfusion that would be predicted given only the underlying anatomy and a residual image that represents perfusion information that cannot be predicted by anatomical features. Our learned structural features are significantly better at predicting brain perfusion than tissue probability maps, which are the input to standard partial volume correction techniques. Studies in test-retest data show that both the anatomically predicted and residual perfusion signal are highly replicable for a given subject. In a pediatric population, both the raw perfusion and structurally predicted images are tightly linked to age throughout adolescence throughout the brain. Interestingly, the residual perfusion also shows a strong correlation with age in select regions including the hippocampi (corr= 0.38, p-value < 10−6), precuneus (corr= −0.44, p < 10−5), and combined default mode network regions (corr= −0.45, p < 10−8) that is independent of global anatomy-perfusion trends. This finding suggests that there is a regionally heterogeneous pattern of functional specialization that is distinct from that of cortical structural development.

Arterial spin labeling perfusion predicts longitudinal decline in semantic variant primary progressive aphasia

“…It is unlikely that there is normal perfusion in these areas of GM thinning since hypoperfusion was extensive in these areas in uncorrected CBF maps (supplementary figure e-2), and these regions have previously been shown to have PET glucose hypometabolism in patients with svPPA [4, 5, 7]. Nonetheless, the partial volume correction approach used in this study to examine structure-function relationships in patients with svPPA relied on an assumed ratio between GM and WM CBF, and future studies should examine other methods of partial volume correction [15]. …”

“…We removed the last 10 images in each series to minimize a previously reported frequency-dependent lipid artifact associated with a similar pCASL sequence [30]. The remaining pCASL images were processed using the antsASLProcessing script in ANTsR, as previously published [15]. Briefly, each pCASL volume was motion-corrected to the mean pCASL image.…”

“…Moreover, ASL has advantages over PET methods for measuring brain function, including functional and structural imaging in the same scanning session, the lack of ionizing radiation associated with PET, and lower cost [11]. ASL has previously been used to investigate neurodegeneration in patients with MCI, AD, and FTD [12–15]. However, the relative utility of ASL MRI for predicting longitudinal cortical alterations has yet to be assessed.…”

“…However, the relative utility of ASL MRI for predicting longitudinal cortical alterations has yet to be assessed. Past work has explored the importance of examining the differences, as well as the similarities, between functional and structural findings [15, 16]. Areas of functional hypoperfusion corresponding to thinning in cortical structures may reflect co-occurring functional and structural decline of a neuronal population.…”

Volumetric Image Registration From Invariant Keypoints

“…This matrix is well-known in computer vision, especially with regard to corner detection. Kandel et al used these eigenvectors to align pairs of image patches [16]. In this work, we use them to derive an orientation local to each keypoint, obviating the need for pairwise alignment.…”