Elsevier

Resuscitation

Volume 84, Issue 10, October 2013, Pages 1393-1399
Resuscitation

Clinical paper
Low apparent diffusion coefficient cluster-based analysis of diffusion-weighted MRI for prognostication of out-of-hospital cardiac arrest survivors

https://doi.org/10.1016/j.resuscitation.2013.04.011Get rights and content

Abstract

Objective

Recent studies suggested quantitative analysis of diffusion-weighted magnetic resonance imaging as a promising tool for early prognostication of cardiac arrest patients. However, most of their methods involve significant manual image handling often subjective and difficult to reproduce. Therefore developing a computerized analysis method using easy-to-define characteristics would be useful.

Methods

Comatose out-of-hospital cardiac arrest (OHCA) patients who underwent brain MRI between January 2008 and July 2012 were identified from an OHCA registry. Apparent diffusion coefficient (ADC) axial images were analyzed using a program to detect and characterize clusters of low ADC pixels from six brain regions including frontal, occipital, parietal, rolandic and temporal and basal ganglia region. Identified clusters were ranked according to size, mean ADC and minimum ADC to assess the regional maximum cluster size (MCS), lowest mean ADC (LMEAN) and lowest minimum ADC (LMIN). Their power to predict poor outcome, defined as 6-month CPC 3 or higher, was assessed by contingency table analyses.

Results

51 OHCA patients were eligible during the study period. The sensitivities of MCS, LMEAN and LMIN to detect poor outcome varied according to brain region from 62.5 to 90.0%, 50.0 to 72.5% and 42.5 to 82.5% with their specificities set to 100%, respectively. The MCS of occipital region showed most favorable test profile (sensitivity 90%, specificity 100%; AUROC 0.940, 95% confidence interval 0.874–1.000).

Conclusion

The cluster-based computerized image analysis might be a simple but useful method for prediction of poor neurologic outcome. Future studies validating its prognostic performance are required.

Introduction

Cardiac arrest carries very poor prognosis in terms of both short-term survival and long-term neurologic recovery. Although some do survive and achieve good recovery, most of survivors do not regain consciousness and need prolonged, if not permanent, inpatient care.1 Therefore, accurate prediction of prognosis is often requested, especially in current situation of limited health care resources. However, accurate prediction of neurologic outcome has been difficult task. And it became even more so since widespread adoption of therapeutic hypothermia which made the validity of previously suggested methods doubtful.2, 3, 4

Recent studies had reported quantitative diffusion-weighted imaging (DWI) analysis as a promising tool for prediction of neurologic outcome after cardiac arrest.5, 6, 7, 8 They employed various anatomy or lesion-based quantitative apparent diffusion coefficient (ADC) measurement for the prognostication. Their reported sensitivity for poor outcome was variable ranging from 41 to 93% when their specificity was 100%. In our previous study, we reported that one of such methods could provide useful prognostic information even in patients treated with therapeutic hypothermia.8

However, most of these methods require significant human effort such as outlining regions of interests (ROIs) which can lead to poor reproducibility and generalizability. Therefore, it might be useful if there is a simple, objective but accurate computerized algorithm-based prediction method available. We hypothesized that analysis of locally connected low ADC pixels (low-ADC cluster) identified by computerized algorithm would provide a simple but reproducible starting point for measuring various ADC features. Therefore, the aims of the current study were (1) implementation of the cluster-based image processing and (2) assessment and comparison of various ADC features to find most suitable one for clinical use.

Section snippets

Study setting

This study is based on a prospectively collected registry of emergency department (ED) OHCA patients treated at a 950-bed tertiary academic hospital located in a city with a population of 480,000. The majority of the emergency medical services in this area are provided by the government and its level of care is primarily restricted to basic life support.9 The management of cardiac arrest was based on the recommendations of the 2005 American Heart Association cardiopulmonary resuscitation

Baseline characteristics

Of the 596 OHCA patients who were older than fifteen, 333 patients had return of spontaneous circulation (ROSC). Among them, a total of 56 comatose OHCA patients were evaluated with diffusion-weighted MRI for prognostication. After exclusion of five patients whose brain MRI was done too early (within 12 h after ROSC) or with poor baseline neurology, a total of 51 patients were included for this study (Table 1). The median age of the study population was 63 years (IQR, 42–72), and 11 (21.6%) of

Discussion

In current study, size, mean and minimum ADC of most dominant cluster were significantly associated with long-term neurologic outcome. Their prognostic performances were varied according to cluster attributes and brain regions. Several indicators, especially MCS of occipital region, were found to have better prognostic performance. Our findings suggest that analysis of low-ADC clusters using simple computerized algorithm can provide useful prognostic information.

Accurate prognostication of

Conclusions

The cluster-based computerized image analysis might be a simple but useful method for prediction of poor neurologic outcome. Future studies validating its prognostic performance are required.

Conflicts of interest statement

All authors do not have any conflicts of interest.

Acknowledgment

This study was supported partially by Seoul National University Bundang Hospital (SNUBH) grant 02-2013-135.

References (17)

There are more references available in the full text version of this article.

Cited by (31)

  • Prognosis of Cardiac Arrest—Peri-arrest and Post-arrest Considerations

    2023, Emergency Medicine Clinics of North America
  • European Resuscitation Council and European Society of Intensive Care Medicine Guidelines 2021: Post-resuscitation care

    2021, Resuscitation
    Citation Excerpt :

    These include lowest minimum or mean ADC,400 mean ADC,401 the proportion of brain volume below a given ADC threshold,401,402 and the maximum size of the MRI clusters with minimum ADC.400 Most of these studies assessed global ADC, while one of them assessed regional ADC.400 In all these studies, an ADC threshold for 100% specificity was identified, often with sensitivities above 50%.

  • Prognostication of cardiac arrest survivors using low apparent diffusion coefficient cluster volume

    2016, Resuscitation
    Citation Excerpt :

    However, their credibility has been challenged since the widespread adoption of therapeutic hypothermia which often requires the administration of sedatives and paralytics to the patients whose cellular and drug metabolism are often impaired.6,7 Recently, brain magnetic resonance imaging (MRI), more specifically diffusion-weighted (DW) imaging, has shown promising results for the long-term prognostication of comatose CA patients.8–17 The methods are very diverse, but the most common approaches include use of a scoring system or apparent diffusion coefficient (ADC) sampling from hand-drawn regions of interest (ROI) that requires subjective human interpretation of lesion characteristics or brain anatomy.

View all citing articles on Scopus

A Spanish translated version of the summary of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2013.04.011.

View full text