Automated brain computed tomographic densitometry of early ischemic changes in acute stroke

Abstract

The Alberta Stroke Program Early CT score (ASPECTS) scoring method is frequently used for quantifying early ischemic changes (EICs) in patients with acute ischemic stroke in clinical studies. Varying interobserver agreement has been reported, however, with limited agreement. Therefore, our goal was to develop and evaluate an automated brain densitometric method. It divides CT scans of the brain into ASPECTS regions using atlas-based segmentation. EICs are quantified by comparing the brain density between contralateral sides. This method was optimized and validated using CT data from 10 and 63 patients, respectively. The automated method was validated against manual ASPECTS, stroke severity at baseline and clinical outcome after 7 to 10 days (NIH Stroke Scale, NIHSS) and 3 months (modified Rankin Scale). Manual and automated ASPECTS showed similar and statistically significant correlations with baseline NIHSS ([Formula: see text] and [Formula: see text], respectively) and with follow-up mRS ([Formula: see text] and [Formula: see text]), except for the follow-up NIHSS. Agreement between automated and consensus ASPECTS reading was similar to the interobserver agreement of manual ASPECTS (differences [Formula: see text] point in 73% of cases). The automated ASPECTS method could, therefore, be used as a supplementary tool to assist manual scoring.

Keywords: ASPECTS scoring; computed tomography; densitometry; image processing; stroke.

Fig. 1

Flow chart of the automated brain densitometry method based on ASPECTS. (a) The spatial transformation $T(x)$ matching the atlas image with the patient image is computed. (b) The atlas-label image, $L_{\text{atlas}}(x)$, is transformed by $T(x)$ to produce a labeled image $L_{\text{pat}}(x)$. (c) The atlas is applied as a mask over the patient image. (d) The density distribution is calculated for each ASPECTS region in the left and right hemisphere; $H_L$ and $H_R$, respectively. (e) For each ASPECTS region, i, the brain density shift (${\mathrm{BDS}}_i$) is calculated between contralateral sides. (f) The EIC are detected by thresholding the BDS values, while accounting for irrelevant defects.

Fig. 2

Capture of the graphical interface of the automated brain densitometry method based on the ASPECTS regions. The top row represents the different BDS measures per ASPECTS regions. The bottom rows are visualizations of the density distributions of the ASPECTS regions. From left to right, top to bottom: caudate nucleus; anterior and posterior part of internal capsule (for the ultimate quantification, these two regions were merged into one region, conform manual ASPECTS); lentiform nucleus; insular ribbon; M1; M2; M3; M4; M5, and M6. Red (gray) curves represent histograms of the left hemisphere and yellow (white) curves of the right hemisphere. A leftward shift of the yellow curve towards the lower density values indicates a hypodensity in the right hemisphere. The internal capsula area was divided in an anterior and a posterior part, to ensure detection of EICs when only a single part of the capsula was affected.

Fig. 3

Box plots of the residual registration errors in the training set of 10 patients, showing the effect of the use of brain masks. The left column shows the residual distances to the ventricle contours and the right column the distances to the brain contours.

Fig. 4

Bland-Altman plots of (a) auto-ASPECTS versus consensus-ASPECTS and (b) differences between observers. The size of the circles and the value inside indicate the number of cases with the same result.

Fig. 5

Percent agreement between observers (cylinders) and between automated and consensus scoring (blocks), for each ASPECTS region. The lower part of the columns represents the percentage agreement on positive EIC findings (i.e., the percentage of cases where both methods indicate a positive EIC finding), the upper part represents the negative agreement (% agreement of a negative EIC finding). From left to right: caudate nucleus (C); internal capsule (IC); lentiform nucleus (L); insular ribbon (I); M1; M2; M3;M4, M5 and M6.