Apparent diffusion coefficient signature of ischemic tissue predicts neurological progression in isolated pontine infarcts

Abstract

Background

Early neurological deterioration is encountered in up to a third of patients with isolated pontine infarcts. A limited number of clinical and imaging features have been suggested as predictors of neurological progression in this setting. In this study, we assessed whether quantitative apparent diffusion coefficient (ADC) measurements within the ischemic pontine region could be used as a radiomic feature to forecast clinical deterioration.

Methods

We calculated the mean ADC value of ischemic voxels within the ischemic region and normalized them to the contralateral non-ischemic tissue (relative ADC, rADC) in patients with isolated pontine infarcts. This imaging signature was then compared among patients with neurological progression (n = 21) and a propensity matched cohort of non-progressors (n = 42), together with other clinical and imaging features in bivariate and multivariate statistical models.

Results

The rADC_mean was significantly lower among patients with progression (p = 0.008). Female gender and extension of the ischemic lesion to the ventral pontine surface were other features significantly associated with progression. The association between rADC_mean and progression persisted in multivariate models with an odds ratio of 13.7 (95% CI 2.6–72.8; p = 0.002) for progression among patients with rADC_mean ≤ 0.67 in their ischemic tissue. The probability for worsening was 80% among patients who had an ischemic lesion extending to the ventral pontine surface with a mean rADC ≤ 0.67.

Conclusion

The mean rADC value within the ischemic lesion is closely related with early neurological deterioration in patients with isolated pontine infarcts.

Introduction

Isolated pontine infarctions constitute approximately 15% of all posterior circulatory ischemic strokes. ¹ Although various etiologies including atrial fibrillation, basilar artery dissection, dolichoectasia, or stenosis could underlie the syndrome in occasional patients, the majority are secondary to an arterial pathology directly involving the basilar penetrator arteries. ² In this regard, two distinct pathological mechanisms have been highlighted based on pathological and radiological studies; the smaller, lacunar type pontine infarctions which generally lie deep in the pons and occur due to lipohyalinosis of the penetrator arteries, and the relatively larger infarctions due to atheromatous branch occlusion of penetrator origins which classically lead to wedge-shaped lesions involving the ventral pontine surface primarily in the paramedian region.^{3, 4} In addition to this mechanistic diversity, one additional challenge in the care of patients with acute pontine infarcts is the tendency of clinical deficits to fluctuate or progress in the first few days after symptom onset, with reports highlighting early motor deterioration ranging from 19% to 32% of cases.^5–11

The determination of high-risk patients for progression is important as these patients are well-known to have worse neurological outcomes in the long term.^8,12 Female sex, history of hypertension and diabetes mellitus, and higher admission National Institutes of Health Stroke Scale (NIHSS) scores have been highlighted as clinical factors that might be predictive of neurological worsening.^7,8,10,11,13 Several other studies underlined the importance of radiological features in predicting progression and suggested caudal location, longer vertical extent, and a pattern suggestive of branch artery occlusion as lesion characteristics associated with a higher rate of early motor deterioration.^6,8,10,13 The magnitude of the decrease in apparent diffusion coefficient (ADC) is closely related to fate of clinical and tissue outcome in territorial strokes.^14–16 In this study, our aim was to assess whether a similar ADC based approach could be used as an additional radiomic feature to forecast clinical deterioration in acute pontine infarctions.

Methods

This was a retrospective evaluation of adult stroke patients admitted to our acute units with magnetic resonance imaging (MRI) confirmed isolated pontine infarctions between September 2009 and December 2020. Patients were identified from the prospectively collected institutional stroke database. The analyses were restricted to patients who had an MRI within 48 hours of symptom onset. Patients were excluded if neurological progression occurred after admission, but prior to obtaining MRI. Neurological progression was defined as an increase in NIHSS arm or leg motor scores ≥1 with respect to baseline within 72 hours after imaging, which is a prospectively collected information in our database where any worsening in neurological symptoms is systematically recorded. The study was approved by the institutional review board.

We extracted age, gender, stroke risk factors, and admission NIHSS score information from the database for all patients. Patients with ≥ 50% stenosis in the basilar artery on angiographic studies were considered to have significant parent artery pathology. All imaging studies were performed on 1.5 T MR scanners (Magnetom Symphony and Magnetom Aera, Siemens Healthineers, Erlangen, Germany and Achieva, Philips, the Netherlands). Diffusion-weighted imaging (DWI) was obtained by single shot echo planar imaging applied with three b values with a maximum of 1000 s/mm² and a slice thickness of 5 mm in all scanners. Repetition time (TR) ranged between 2600 and 7100 ms, and echo time (TE) ranged between 87 and 116 ms across the three scanners. For ADC calculations, the ischemic area was manually segmented on DWI using 3D-Slicer version 4.11 (http://www.slicer.org). ¹⁷ These segments were then exported and overlayed onto the apparent diffusion coefficient (ADC) maps of the same patient. After checking for proper positioning of each segmentation on its respective ADC map, an ADC signal intensity histogram was obtained using the Python interface available on the software. The histogram was then used to obtain the mean and standard deviation of the ADC values within the voxels of the ischemic lesion (Figure 1). The mean ADC value of the ischemic voxels was then normalized to the ADC values of the mirror region of interest in the contralateral, non-ischemic side of the pons and relative ADC values (rADC) were obtained. The extension of the lesion to the ventral pontine surface was evaluated on trace images. All radiological assessments were performed while blinded to the neurological outcome.

Figure 1.

Demonstrative figure summarizing methodology for image analysis. Acute ischemic lesions on DWI (A) were manually outlined to create lesion masks (green) (B; top row). The lesion masks were then overlayed to the corresponding ADC maps (B; second row). A second mask (yellow) was automatically created on the contralateral non-ischemic pontine area mirroring the ischemic region (C). Accurate positioning of the mirror mask was also verified in coronal planes (D) to avoid partial volume artifacts arising from the cerebrospinal fluid. Both masks were then used to create ADC histograms in the ischemic (E) and contralateral (F) pontine tissue.

Statistical Analysis:

Categorical variables are expressed as n (%) and continuous variables as median (interquartile range, IQR). Propensity matching with a 1:2 ratio and based on age and time from symptom onset to MRI was used to identify a non-progressor cohort to be used in analyses. Chi-square test and Mann–Whitney U tests were used to evaluate group wise differences. Logistic regression analysis was performed to identify independent predictors of neurological progression. A p value of < 0.05 was considered statistically significant. The analyses were done using IBM SPSS Statistics for Macintosh, version 23.0 (Armonk, NY: IBM Corp).

Results

During the study period, a total of 212 patients were admitted with acute ischemic lesions solely localized to the pontine area. Of these, MRI was available within 48 hours of symptom onset in 150 patients; however, we excluded an additional four cases with neurological progression prior to MRI. Of the remaining 146 patients, early neurological progression occurred in 21 patients (14%). A total of 42 patients from the group of non-progressors were selected by propensity matching as the comparator group.

Table 1 summarizes the clinical and radiological characteristics of the study population. Patients showing neurological progression were more likely to be female and harbor ischemic lesions extending to the ventral surface of pons. No significant difference was evident in terms of stroke risk factors, stroke severity, or treatments applied prior to progression. The rADC_mean was significantly lower among patients with progression (p = 0.008). The AUC of the ROC curve for rADC_mean to classify patients with and without progression was 0.71 (95% CI 0.57–0.85) with an optimal operating point at the value of rADC_mean = 0.67 (sensitivity 91% and specificity 48%).

Table 1:

Clinical and Radiological Features Among Study Groups

	Neurological Progression		p
	Yes (n = 21)	No (n = 42)
Age (year)	68 (63–76)	69 (65–76)	0.715
Female gender	11 (52%)	10 (24%)	0.023
Hypertension	20 (95%)	33 (79%)	0.144
Diabetes mellitus	12 (57%)	17 (41%)	0.211
Hyperlipidemia	13 (62%)	25 (60%)	0.856
Coronary artery disease	7 (33%)	14 (33%)	1.000
Atrial fibrillation	3 (14%)	2 (5%)	0.323
Prior stroke	9 (43%)	12 (29%)	0.257
Admission NIHSS score	4 (1–6)	3 (1–5)	0.449
Time from symptom onset to MRI (hr)	13 (9–19)	14 (8–23)	0.710
iv-tPA therapy prior to progression	1 (5%)	2 (5%)	1.000
Dual antiplatelet therapy prior to progression	6 (29%)	10 (24%)	0.682
Parent artery disease	5 (24%)	6 (14%)	0.348
Extension of ischemic lesion to the ventral pontine surface	19 (91%)	26 (62%)	0.020
Admission DWI lesion volume (mL)	0.7 (0.4–1.0)	0.6 (0.3–0.9)	0.466
Relative ADCmean of ischemic voxels	0.69 (0.63–0.79)	0.78 (0.71–0.84)	0.008

All categorical variables are expressed as n (%) and continuous variables as median (IQR)

A multivariate model including female gender, lesion topography, and rADC_mean showed extension of lesions to ventral pontine surface (OR 10.4, 95% CI 1.5–72.1; p = 0.018) and rADC_mean ≤ 0.67 (OR 13.7, 95% CI 2.6–72.8; p = 0.002) as independent predictors of neurological progression. The neurological exam showed no evidence of worsening among patients without any of these imaging criteria (n = 14), while progression was observed in 33% of patients with one imaging criteria (n = 39) and in 80% of those with both criteria present (n = 10).

Discussion

The study underlines the importance of rADC within the ischemic lesion as an additional radiological parameter from the perspective of neurological progression in pontine infarcts. A mean ADC in the ischemic region less than two-thirds of the contralateral non-ischemic tissue was closely associated with progression of neurological deficits over the ensuing days after stroke. This association was independent of previously reported variables related with a detrimental clinical course, such as female gender or ventral extension of the DWI lesion to the pontine surface.

The prediction of neurological progression is challenging in penetrator artery occlusions when compared to larger, territorial strokes where risk stratification can be performed by a multitude of readily available radiological tools. Despite some promising publications, the small size of the lesion, together with the technical difficulties of perfusion imaging in posterior fossa, significantly hinder the utility of such a tissue-based prediction algorithm in pontine infarcts.^18,19 Likewise, routinely used angiographic tools which can only be used to assess the parent vessel are not informative regarding vascular status or collateral supply in the ischemic area. Moreover, the presence of significant stenosis in the parent artery is not predictive of neurological outcome as shown in the previous literature, and also in our study.^6,10 Although the prognostic importance of certain clinical features, including gender or vascular risk factors, have been highlighted in pontine infarcts, their utility in terms of predicting motor deterioration is extremely limited. This leaves the clinician with a limited number of tools to tease out patients with a high risk of worsening from those with favorable prognosis.^6,8,10

Quantitative assessment of ADC maps is a well-known imaging tool that provides prognostic information in non-lacunar strokes. Lower values of ADC are not only predictive of the irreversibility of the diffusion lesion but also are considered an ominous sign of infarct growth in the penumbral tissue.^14–16,20 Apart from the mean value of ADC in the region of interest, the voxel-wise heterogeneity seems to be important as well and can signify a potential for tissue or neurological recovery. ²¹ Finally, from the perspective of small vessel disease, lower ADC values at admission were shown to be predictive of lacunarization on follow-up studies, possibly as a reflection of the severity of the ischemic insult. ²² All these findings make quantitative ADC evaluation an appealing target for predicting neurological worsening in patients with isolated pontine infarcts, especially in the scarcity of other radiological tools. Our study is an initial effort in this regard and shows that lower ADC values in ischemic voxels are highly predictive of early neurological deterioration in these patients.

Our study has a number of limitations. First of all, the ADC threshold used in the analyses has not been validated in an external cohort and needs to be evaluated in independent datasets in terms of sensitivity and specificity. In addition, images were obtained on multiple MRI scanners with similar magnetic field strength, yet with a certain amount of heterogeneity in acquisition parameters other than b value and slice thickness. We tried to minimize the potential variability in ADC values originating from different acquisition protocols by normalizing the measurements within the ischemic territory to the contralateral non-ischemic area. Furthermore, there was no significant difference in TR and TE values among patients with or without early neurologic progression. Still, our results need to be replicated in larger cohorts with use of standardized imaging protocols, which would not only be critical for understanding the physiological underpinnings of neurological progression in patients with isolated pontine infarcts but also for development of tools to identify patients with high risk of progression in the early hours of stroke, where more aggressive therapeutic approaches could be tested.

ORCID iDs

Dogan Dinc Oge https://orcid.org/0000-0001-8103-4779

Ethem Murat Arsava https://orcid.org/0000-0002-6527-4139