Limits...
The walking-induced transient hack concept is valid & relies on a transient early-exercise hypoxemia.

Bruneau A, Feuilloy M, Dussaussoy C, Gagnadoux F, Leftheriotis G, Abraham P - PLoS ONE (2013)

Bottom Line: Clustering classes were compared to the profile types previously proposed with the cross-correlation technique.The classifications of patients according to both approaches were compared using kappa statistics.Clustering analysis resulted in 4 classes that closely fit the 4 most frequently proposed empirical models (cross-correlation coefficients: 0.93 to 0.97).

View Article: PubMed Central - PubMed

Affiliation: L'Université Nantes Angers Le Mans, Centre Hospitalier Universitaire d'Angers, Laboratory for Vascular Investigations, Angers, France.

ABSTRACT

Background: Decreased arterial oxygen pressure obtained at peak exercise is strong evidence of walking-induced hypoxemia, assuming that the lower pressure occurs just before exercise is stopped. Using empirical predefined models and transcutaneous oximetry, we have shown that some patients reporting exercise intolerance show a minimal value at the onset of walking and a post-exercise overshoot. These changes are referred to as transcutaneous "walking-induced transient hacks".

Methods: In 245 patients, walking-induced transcutaneous oxygen pressure changes in the chest were analyzed using observer-independent clustering techniques. Clustering classes were compared to the profile types previously proposed with the cross-correlation technique. The classifications of patients according to both approaches were compared using kappa statistics. In 10 patients showing a hack on transcutaneous oximetry, we analyzed the results of direct iterative arterial sampling recorded during a new walking treadmill test.

Results: Clustering analysis resulted in 4 classes that closely fit the 4 most frequently proposed empirical models (cross-correlation coefficients: 0.93 to 0.97). The kappa between the two classifications was 0.865. In 10 patients showing transcutaneous hacks, the minimal direct arterial oxygen pressure value occurred at exercise onset, and these patients exhibited a recovery overshoot reaching a maximum at two minutes of recovery, confirming the walking-induced transient hypoxemia.

Conclusions: In patients reporting exercise intolerance, transcutaneous oximetry could help to detect walking-induced transient hypoxemia, while peak-exercise arterial oximetry might be normal.

Show MeSH

Related in: MedlinePlus

Comparison between the empirical models and the statistical models resulting from the clustering analysis.Empirical models are dotted lines, and statistical models are solid lines. In the figure, the exercise period is in gray. In the table, the highest coefficients are in bold characters.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3643941&req=5

pone-0062651-g002: Comparison between the empirical models and the statistical models resulting from the clustering analysis.Empirical models are dotted lines, and statistical models are solid lines. In the figure, the exercise period is in gray. In the table, the highest coefficients are in bold characters.

Mentions: Thereafter, merging of initial relevant clusters into classes was achieved by hierarchical agglomerative clustering (HAC). This approach determines successive levels of classes from an initial set of pre-established clusters until obtaining a single class. At each level, it organizes the classes by merging the two most similar clusters or classes. This method organizes the classes into a hierarchical structure according to a proximity matrix. The proximity matrix represents the similarity measures between all pairs of clusters. The similarity index constructing the classes depends on the distances between the symbolic representations of the previous level’s clusters and is based on a rank correlation index. The process of hierarchical clustering is represented in figure 2 by a tree structure called a dendrogram: it shows how the clusters have been grouped together level by level. The partition of the final classes is obtained by choosing the cutoff on the dendrogram. To optimize the cutoff, we computed a score based on the ratio of the between-class variance to the within-class variance for each step of the HAC process. This score can be interpreted like Fisher’s criterion [16], where the optimal partition maximizes and minimizes, respectively, the between-class and within-class variances. This resulted in 4 mathematically defined classes A′, B′, C′ and D′. Within each class, a curve was obtained by averaging all waveforms belonging to each patient of the class.


The walking-induced transient hack concept is valid & relies on a transient early-exercise hypoxemia.

Bruneau A, Feuilloy M, Dussaussoy C, Gagnadoux F, Leftheriotis G, Abraham P - PLoS ONE (2013)

Comparison between the empirical models and the statistical models resulting from the clustering analysis.Empirical models are dotted lines, and statistical models are solid lines. In the figure, the exercise period is in gray. In the table, the highest coefficients are in bold characters.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3643941&req=5

pone-0062651-g002: Comparison between the empirical models and the statistical models resulting from the clustering analysis.Empirical models are dotted lines, and statistical models are solid lines. In the figure, the exercise period is in gray. In the table, the highest coefficients are in bold characters.
Mentions: Thereafter, merging of initial relevant clusters into classes was achieved by hierarchical agglomerative clustering (HAC). This approach determines successive levels of classes from an initial set of pre-established clusters until obtaining a single class. At each level, it organizes the classes by merging the two most similar clusters or classes. This method organizes the classes into a hierarchical structure according to a proximity matrix. The proximity matrix represents the similarity measures between all pairs of clusters. The similarity index constructing the classes depends on the distances between the symbolic representations of the previous level’s clusters and is based on a rank correlation index. The process of hierarchical clustering is represented in figure 2 by a tree structure called a dendrogram: it shows how the clusters have been grouped together level by level. The partition of the final classes is obtained by choosing the cutoff on the dendrogram. To optimize the cutoff, we computed a score based on the ratio of the between-class variance to the within-class variance for each step of the HAC process. This score can be interpreted like Fisher’s criterion [16], where the optimal partition maximizes and minimizes, respectively, the between-class and within-class variances. This resulted in 4 mathematically defined classes A′, B′, C′ and D′. Within each class, a curve was obtained by averaging all waveforms belonging to each patient of the class.

Bottom Line: Clustering classes were compared to the profile types previously proposed with the cross-correlation technique.The classifications of patients according to both approaches were compared using kappa statistics.Clustering analysis resulted in 4 classes that closely fit the 4 most frequently proposed empirical models (cross-correlation coefficients: 0.93 to 0.97).

View Article: PubMed Central - PubMed

Affiliation: L'Université Nantes Angers Le Mans, Centre Hospitalier Universitaire d'Angers, Laboratory for Vascular Investigations, Angers, France.

ABSTRACT

Background: Decreased arterial oxygen pressure obtained at peak exercise is strong evidence of walking-induced hypoxemia, assuming that the lower pressure occurs just before exercise is stopped. Using empirical predefined models and transcutaneous oximetry, we have shown that some patients reporting exercise intolerance show a minimal value at the onset of walking and a post-exercise overshoot. These changes are referred to as transcutaneous "walking-induced transient hacks".

Methods: In 245 patients, walking-induced transcutaneous oxygen pressure changes in the chest were analyzed using observer-independent clustering techniques. Clustering classes were compared to the profile types previously proposed with the cross-correlation technique. The classifications of patients according to both approaches were compared using kappa statistics. In 10 patients showing a hack on transcutaneous oximetry, we analyzed the results of direct iterative arterial sampling recorded during a new walking treadmill test.

Results: Clustering analysis resulted in 4 classes that closely fit the 4 most frequently proposed empirical models (cross-correlation coefficients: 0.93 to 0.97). The kappa between the two classifications was 0.865. In 10 patients showing transcutaneous hacks, the minimal direct arterial oxygen pressure value occurred at exercise onset, and these patients exhibited a recovery overshoot reaching a maximum at two minutes of recovery, confirming the walking-induced transient hypoxemia.

Conclusions: In patients reporting exercise intolerance, transcutaneous oximetry could help to detect walking-induced transient hypoxemia, while peak-exercise arterial oximetry might be normal.

Show MeSH
Related in: MedlinePlus