Purpose

Cerebral small vessel disease (CSVD) refers to microvascular brain pathologies leading to various types of tissue lesions. Two CSVD-subtypes are differentiated for treatment of the underlying primary disease: cerebral amyloid angiopathy (CAA) and hypertensive arteriopathy (HA). Differentiation is ambiguous in patients with signs of both types. An analysis of brain lesion load may contribute to reducing this ambiguity.

Materials & Methods

We propose a web-based tool for an interactive visual analysis of individual and cohort CSVD lesion load. The tool currently supports three lesion types: white matter hyperintensities, enlarged perivascular spaces and cerebral microbleeds. In a pre-processing step, the lesions must be segmented in different MRI sequences and the image data as well as the masks must be co-registered. The tool integrates multiple views linked for cohort specification [1] and the visualization of lesion load (Fig. 1). Moreover, it supports the comparison of two subcohorts, e.g., potential CAA and HA cases. A detailed representation of the individual lesions and lesion co-occurrence are shown in 2D and 3D. Since the cohort representation can quickly become cluttered due to many lesions and several intertwining structures, we included a more abstract visualization of the lesion load distribution in different brain regions using the bullseye parcellation [2].

Results

We evaluated the tool in a pilot study with 10 cases from clinical routine. The lesions were segmented using mdbrain (mediaire GmbH) and Mango (RII, UT Health). The tool was tested by three CSVD experts. They stated that the tool is very useful in detecting spatial lesion patterns that may be characteristic for CAA or HA and support a better differentiation of mixed cases. They found that some cases had an asymmetrical lesion load in the frontal lobes, which deserves further investigation.

Discussion

The tool provides new insights about the lesion load and pattern of CSVD patients. To improve the analysis process, additional visualizations highlighting the similarities of two subcohorts as well as a statistical evaluation should be included in the tool. A follow-up study of 240 clinical cases is in progress.

Conclusion

The general principle of an interactive visual analysis of CSVD lesion load is a very promising approach to further research in CAA-HA-differentiation.

Sources

[1] Müller, J., et al. IEEE Trans. Vis. Comput. Graph. 27.6 (2021)

[2] Sudre, C. H., et al. J Neuroradiol. 45.2 (2018)