Contributions
Abstract: P1055
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
Introduction: Whole brain (WB) and grey matter (GM) volume decrease faster in MS patients compared to a healthy population. In order to provide a reference for a patient"s brain volume at a certain time, the brain volume can be compared against volumes of a healthy population. MSmetrix is an MRI-based brain segmentation tool that provides WB and GM volumes, including a comparison against an age and sex matched population. After normalization for head size, the result is a normative percentile. In this work, the stability and thus the usability in clinical practice of the healthy population graphs used in MSmetrix are validated.
Materials and methods: Data included in the population graphs consists of 1281 healthy subjects (761 females, 520 males; ages 18-96) retrieved from publicly available MRI collections. A leave-one-out approach is used, where each subject is consecutively removed from the population and placed onto an independently computed population graph, obtained from the remaining subjects.
Normalized WB and GM volumes computed by MSmetrix are used for constructing three different population graphs, based on the female, male and all datasets, respectively. Validation experiments are designed to test the stability of the population graph for each possible combination: WB or GM for female, male, or all datasets. For each of the combinations, the separated subjects" brain volume is compared to the remaining population to construct a sample of percentiles. A Kolmogorov-Smirnov test was performed to evaluate whether the sample of percentiles computed through this leave-one-out process spans the interval (0, 100) uniformly.
Results: For all population graphs (WB or GM; female, male, or all), the cumulative distribution of the percentiles and true uniform distribution are very close, with a maximum distance below 0.03, and all p-values for the Kolmogorov-Smirnov goodness-of-fit test in the range 0.82 - 0.99. Therefore there is no evidence, at 0.1 significance level (p-value < 0.1), to reject the hypothesis that the percentiles of the leave-one-out subjects would not be uniformly distributed.
Conclusion: The validation experiments indicate that the computation of the population graph is stable, for both normalized WB and normalized GM, as well as for the three types of population graphs (female, male, all). This means the population graphs are a reliable reference for brain volume measurements, which is useful in the clinical follow up of MS patients.
Disclosure: Diana M. Sima: employee of icometrix
Saurabh Jain: employee of icometrix
Anke Maertens: employee of icometrix
Eline Van Vlierberghe: employee of icometrix
Wim Van Hecke: shareholder of icometrix
Dirk Smeets: employee of icometrix
Abstract: P1055
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
Introduction: Whole brain (WB) and grey matter (GM) volume decrease faster in MS patients compared to a healthy population. In order to provide a reference for a patient"s brain volume at a certain time, the brain volume can be compared against volumes of a healthy population. MSmetrix is an MRI-based brain segmentation tool that provides WB and GM volumes, including a comparison against an age and sex matched population. After normalization for head size, the result is a normative percentile. In this work, the stability and thus the usability in clinical practice of the healthy population graphs used in MSmetrix are validated.
Materials and methods: Data included in the population graphs consists of 1281 healthy subjects (761 females, 520 males; ages 18-96) retrieved from publicly available MRI collections. A leave-one-out approach is used, where each subject is consecutively removed from the population and placed onto an independently computed population graph, obtained from the remaining subjects.
Normalized WB and GM volumes computed by MSmetrix are used for constructing three different population graphs, based on the female, male and all datasets, respectively. Validation experiments are designed to test the stability of the population graph for each possible combination: WB or GM for female, male, or all datasets. For each of the combinations, the separated subjects" brain volume is compared to the remaining population to construct a sample of percentiles. A Kolmogorov-Smirnov test was performed to evaluate whether the sample of percentiles computed through this leave-one-out process spans the interval (0, 100) uniformly.
Results: For all population graphs (WB or GM; female, male, or all), the cumulative distribution of the percentiles and true uniform distribution are very close, with a maximum distance below 0.03, and all p-values for the Kolmogorov-Smirnov goodness-of-fit test in the range 0.82 - 0.99. Therefore there is no evidence, at 0.1 significance level (p-value < 0.1), to reject the hypothesis that the percentiles of the leave-one-out subjects would not be uniformly distributed.
Conclusion: The validation experiments indicate that the computation of the population graph is stable, for both normalized WB and normalized GM, as well as for the three types of population graphs (female, male, all). This means the population graphs are a reliable reference for brain volume measurements, which is useful in the clinical follow up of MS patients.
Disclosure: Diana M. Sima: employee of icometrix
Saurabh Jain: employee of icometrix
Anke Maertens: employee of icometrix
Eline Van Vlierberghe: employee of icometrix
Wim Van Hecke: shareholder of icometrix
Dirk Smeets: employee of icometrix