Application of principal component analysis to study topography of hypoxic–ischemic brain injury

Singhal, Shaloo; Chen, Jian; Beare, Richard; Ma, Henry; Ly, John; Phan, Thanh G.

doi:10.1016/j.neuroimage.2012.04.025

Cited by 4 publications

(5 citation statements)

References 41 publications

(60 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The limitations of this study have been described in a related paper on topography of hypoxic ischemic injury ( 10 ). These limitations include retrospective nature, small sample size (only 13% of the comatosed patient had MR scans), and potential for underestimation of white matter ischemic injury when using conventional imaging.…”

Section: Discussionmentioning

confidence: 99%

“…The data here have been discussed in our previous study, which compared methodologies for comparison of topographic imaging findings, rather than outcome prediction ( 10 ).…”

Section: Methodsmentioning

confidence: 99%

“…The imaging method was described in a previous paper from our group ( 10 ). In brief, infarct tissue was manually segmented on T 2 -weighted and B1000 images of the DWI sequence.…”

Section: Methodsmentioning

confidence: 99%

See 2 more Smart Citations

Exploratory Use of Decision Tree Analysis in Classification of Outcome in Hypoxic–Ischemic Brain Injury

et al. 2018

Self Cite

View full text Add to dashboard Cite

BackgroundPrognostication following hypoxic ischemic encephalopathy (brain injury) is important for clinical management. The aim of this exploratory study is to use a decision tree model to find clinical and MRI associates of severe disability and death in this condition. We evaluate clinical model and then the added value of MRI data.MethodThe inclusion criteria were as follows: age ≥17 years, cardio-respiratory arrest, and coma on admission (2003–2011). Decision tree analysis was used to find clinical [Glasgow Coma Score (GCS), features about cardiac arrest, therapeutic hypothermia, age, and sex] and MRI (infarct volume) associates of severe disability and death. We used the area under the ROC (auROC) to determine accuracy of model. There were 41 (63.7% males) patients having MRI imaging with the average age 51.5 ± 18.9 years old. The decision trees showed that infarct volume and age were important factors for discrimination between mild to moderate disability and severe disability and death at day 0 and day 2. The auROC for this model was 0.94 (95% CI 0.82–1.00). At day 7, GCS value was the only predictor; the auROC was 0.96 (95% CI 0.86–1.00).ConclusionOur findings provide proof of concept for further exploration of the role of MR imaging and decision tree analysis in the early prognostication of hypoxic ischemic brain injury.

show abstract

Section: Discussionmentioning

confidence: 99%

“…The data here have been discussed in our previous study, which compared methodologies for comparison of topographic imaging findings, rather than outcome prediction ( 10 ).…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Exploratory Use of Decision Tree Analysis in Classification of Outcome in Hypoxic–Ischemic Brain Injury

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…These covariance based methods involve regressing the voxels in the images and the outcome of interest. One of the disadvantages with both principal component regression [ 27 ] and partial least squares [ 13 , 14 , 28 , 29 ] that we identified is the inability of these methods to take into account other covariates. This effect is likely to be due to the use of voxel type analysis compared to multiple regions of interests approach (eg.…”

Section: Discussionmentioning

confidence: 99%

Googling Stroke ASPECTS to Determine Disability: Exploratory Analysis from VISTA-Acute Collaboration

et al. 2015

Self Cite

View full text Add to dashboard Cite

The summed Alberta Stroke Program Early CT Score (ASPECTS) is useful for predicting stroke outcome. The anatomical information in the CT template is rarely used for this purpose because traditional regression methods are not adept at handling collinearity (relatedness) among brain regions. While penalized logistic regression (PLR) can handle collinearity, it does not provide an intuitive understanding of the interaction among network structures in a way that eigenvector method such as PageRank can (used in Google search engine). In this exploratory analysis we applied graph theoretical analysis to explore the relationship among ASPECTS regions with respect to disability outcome. The Virtual International Stroke Trials Archive (VISTA) was searched for patients who had infarct in at least one ASPECTS region (ASPECTS ≤9, ASPECTS=10 were excluded), and disability (modified Rankin score/mRS). A directed graph was created from a cross correlation matrix (thresholded at false discovery rate of 0.01) of the ASPECTS regions and demographic variables and disability (mRS>2). We estimated the network-based importance of each ASPECTS region by comparing PageRank and node strength measures. These results were compared with those from PLR. There were 185 subjects, average age 67.5± 12.8 years (55% Males). Model 1: demographic variables having no direct connection with disability, the highest PageRank was M2 (0.225, bootstrap 95% CI 0.215-0.347). Model 2: demographic variables having direct connection with disability, the highest PageRank were M2 (0.205, bootstrap 95% CI 0.194-0.367) and M5 (0.125, bootstrap 95% CI 0.096-0.204). Both models illustrate the importance of M2 region to disability. The PageRank method reveals complex interaction among ASPECTS regions with respects to disability. This approach may help to understand the infarcted brain network involved in stroke disability.

show abstract

“…We achieved this aim by using principal component analysis (PCA) with varimax rotation to investigate the patients’ lesion structure. One previous study on coma used PCA on brain images to compare with a probabilistic frequency map ( Singhal et al , 2012 ). They identified six components in the patients’ data and concluded that the PCA methodology was better equipped to depict patterns of co-varying damage than probabilistic frequency maps.…”

Section: Introductionmentioning

confidence: 99%

Evaluating the granularity and statistical structure of lesions and behaviour in post-stroke aphasia

Zhao

Halai

Ralph

2020

Brain Communications

View full text Add to dashboard Cite

The pursuit of relating the location of neural damage to the pattern of acquired language and general cognitive deficits post-stroke stems back to 19th century behavioural neurology. Whilst spatial specificity has improved dramatically over time, from the large areas of damage specified by post-mortem investigation to the millimetre precision of modern MRI, there is an underlying issue that is rarely addressed, which relates to the fact that damage to a given area of the brain is not random but constrained by the brain’s vasculature. Accordingly, the aim of this study was to uncover the statistical structure underlying the lesion profile in chronic aphasia post-stroke. By applying varimax-rotated principal component analysis to the lesions of 70 patients with chronic post-stroke aphasia, we identified 17 interpretable clusters, largely reflecting the vascular supply of middle cerebral artery sub-branches and other sources of individual variation in vascular supply as shown in classical angiography studies. This vascular parcellation produced smaller displacement error in simulated lesion-symptom analysis compared with individual voxels and Brodmann regions. A second principal component analysis of the patients’ detailed neuropsychological data revealed a four-factor solution reflecting phonological, semantic, executive-demand and speech fluency abilities. As a preliminary exploration, stepwise regression was used to relate behavioural factor scores to the lesion principal components. Phonological ability was related to two components, which covered the posterior temporal region including the posterior segment of the arcuate fasciculus, and the inferior frontal gyrus. Three components were linked to semantic ability and were located in the white matter underlying the anterior temporal lobe, the supramarginal gyrus and angular gyrus. Executive-demand related to two components covering the dorsal edge of the middle cerebral artery territory, while speech fluency was linked to two components that were located in the middle frontal gyrus, precentral gyrus, and subcortical regions (putamen and thalamus). Future studies can explore in formal terms the utility of these principal component analysis derived lesion components for relating post-stroke lesions and symptoms.

show abstract

Application of principal component analysis to study topography of hypoxic–ischemic brain injury

Cited by 4 publications

References 41 publications

Exploratory Use of Decision Tree Analysis in Classification of Outcome in Hypoxic–Ischemic Brain Injury

Exploratory Use of Decision Tree Analysis in Classification of Outcome in Hypoxic–Ischemic Brain Injury

Googling Stroke ASPECTS to Determine Disability: Exploratory Analysis from VISTA-Acute Collaboration

Evaluating the granularity and statistical structure of lesions and behaviour in post-stroke aphasia

Contact Info

Product

Resources

About