Boosting power for clinical trials using classifiers based on multiple biomarkers

“where i is the slack variable for each i (Burges, 1998; Vapnik, 1995).”

“With this modification, the optimization problem may be restated as subject to y i ( x i · w + b ) ≥ 1 − ξ i , where ξ i is the slack variable for each i (Vapnik, 1998, Burges, 1998).”

“We performed our experiments using the LS-SVM package for classification and regression (Suykens and Vandewalle, 1999) in Matlab (MathWorks, Natick, MA).”

“We also considered ApoE genotype (coded as 0, 1 or 2 for the number of E4 alleles), age, sex and BMI, as each can influence AD risk (Corder et al., 1993, Lindsay et al., 2002, Azad et al., 2007, Buchman et al., 2005).”

“ROC curves were compared, to evaluate different classifiers, using a statistical method developed for ROC analysis (Hanley and McNeil, 1983) in the MedCalc Statistical Software (MedCalc, Mariakerke, Belgium).”

“For an introduction to SVM - comparing it to simpler methods such as linear discriminant analysis (LDA) - please see our tutorial (Morra et al., 2009b).”

“…as amyloid and tau deposition, measured by Positron Emission Tomography (PET) and radiotracers that bind to the plaques and tangles in the brain (Klunk et al., 2004; Protas et al., 2010), metabolic decline or perfusion deficits assessed by fluoro-deoxyglucose PET (PET-FDG), brain atrophy on…”

“Biologically, hippocampal atrophy and ventricular enlargement are established manifestations of AD pathology, and the two structures are routinely monitored via MRI for AD clinical trials (Frisoni et al., 2010).”

“(MMSE) – and the CSF biomarker measures were those found to be most promising in pilot studies (Shaw et al., 2009).”

“We also considered ApoE genotype (coded as 0, 1 or 2 for the number of E4 alleles), age, sex and BMI, as each can influence AD risk (Corder et al., 1993, Lindsay et al., 2002, Azad et al., 2007, Buchman et al., 2005).”

“When compared to accuracy results reported by groups such as Vemuri et al. (2008), Klöppel et al. (2008) and Fan et al. (2008), our accuracies may seem a bit low.”

“Brain atrophy rates measured by MRI correlate with the progression of Alzheimer’s disease, and offer baseline and transitional predictive power for diagnosis, making them clinically relevant endpoints for power analysis (Duara et al., 2008, Fox et al., 2000, Jack et al., 2004).”

“PET-FDG numerical summaries were based on a pre-defined temporal lobe ROI (Landau et al., 2009).”

“…only for AD, but also for predicting changes in patients with brain tumors (Lukas et al., 2004), aphasia (Wilson et al., 2009), autism (Ecker et al., 2010), psychosis (Koutsouleris et al., 2009) and even for classifying patterns of brain activation in functional MRI (Mourão-Miranda et al., 2005).”

“BMI was included as several recent studies found that higher BMI is associated with greater brain atrophy in normal elderly subjects (Raji et al., 2010), and in MCI and AD (Ho et al.”

“Other groups report comparable power for measures based on hippocampal volumes (Schuff et al., 2009).”

“Unexpectedly, however, the MRI hippocampal summaries were not incorporated into our predictive model, which is surprising as hippocampal volume can be useful for prediction of MCI progression to AD (Apostolova et al., 2006a; Apostolova et al., 2006b; Apostolova et al., 2007; Frisoni et al., 2010).”

“Vemuri et al. (2008) assigned overall “scores” for each subject’s MRI – called the Structural Abnormality Index (STAND) - based on gray and white matter voxels that best differentiated AD patients from controls.”

“We also considered ApoE genotype (coded as 0, 1 or 2 for the number of E4 alleles), age, sex and BMI, as each can influence AD risk (Corder et al., 1993, Lindsay et al., 2002, Azad et al., 2007, Buchman et al., 2005).”

“Additionally, MRI-based STAND scores were shown to improve the accuracy of CSF biomarkers for predicting cognitive decline, including total tau (t-tau), phosphorylated tau (p-tau) and the beta-amyloid isoform, a β 42 (Vemuri et al., 2009).”

“…learning algorithm for this report due to its successful performance in the previous AD literature (Davatzikos et al., 2009, Fan et al., 2008, Mesrob et al., 2008, Vemuri et al., 2008), and for other neurobiological applications (Ecker et al., 2010, Koutsouleris et al., 2009, Wilson et al., 2009).”

“…learning algorithm for this report due to its successful performance in the previous AD literature (Davatzikos et al., 2009, Fan et al., 2008, Mesrob et al., 2008, Vemuri et al., 2008), and for other neurobiological applications (Ecker et al., 2010, Koutsouleris et al., 2009, Wilson et al., 2009).”

“The temporal lobe summaries were obtained from an anatomically defined region-of-interest (ROI) on three-dimensional atrophy maps generated with tensor-based morphometry (Hua et al., 2008a; Hua et al., 2008b).”

“Brain atrophy rates measured by MRI correlate with the progression of Alzheimer’s disease, and offer baseline and transitional predictive power for diagnosis, making them clinically relevant endpoints for power analysis (Duara et al., 2008, Fox et al., 2000, Jack et al., 2004).”

“MRI-based temporal and ventricular volumes have also been reported for their predictive power in MCI subjects (Fleisher et al., 2008; Korf et al., 2004).”

“The combination of PET-FDG and ApoE genotype has been previously shown to provide good accuracy for predicting MCI conversion (Mosconi et al., 2004).”

“Several studies have performed diagnostic classification based on MRI scans, using measures such as whole-brain patterns of atrophy (Davatzikos et al., 2009; Mesrob et al., 2008), tissue densities from voxel-based morphometry (Vemuri et al., 2008) and cortical thickness (Lerch et al., 2008).”

“As mentioned in the introduction, SVM was chosen as a machine learning algorithm for this report due to its successful performance in the previous AD literature (Davatzikos et al., 2009; Fan et al., 2008; Mesrob et al., 2008; Vemuri et al., 2008), and for other neurobiological applications (Ecker et al.”

“Several studies have performed diagnostic classification based on MRI scans, using measures such as whole-brain patterns of atrophy (Davatzikos et al., 2009; Mesrob et al., 2008), tissue densities from voxel-based morphometry (Vemuri et al., 2008) and cortical thickness (Lerch et al., 2008).”

“By using a multi-modality regression approach to predicting cognitive decline in ADNI subjects, a very recent study found that a linear combination of MRI and PET-FDG was a better predictor of cognitive decline than CSF biomarkers (Walhovd et al., 2010), consistent with our best set of biomarkers.”

“Furthermore, machine learning can perhaps be used to discover genetic (Stein et al., 2009; Stein et al., 2010), epidemiological and physiological factors that influence the progression of AD.”

“Brain atrophy rates measured by MRI correlate with the progression of Alzheimer’s disease, and offer baseline and transitional predictive power for diagnosis, making them clinically relevant endpoints for power analysis (Duara et al., 2008, Fox et al., 2000, Jack et al., 2004).”

“Unexpectedly, however, the MRI hippocampal summaries were not incorporated into our predictive model, which is surprising as hippocampal volume can be useful for prediction of MCI progression to AD (Apostolova et al., 2006a; Apostolova et al., 2006b; Apostolova et al., 2007; Frisoni et al., 2010).”

“Similar algorithms have been implemented to distinguish AD from other types of dementia such as frontotemporal dementia (Davatzikos et al., 2008; Klöppel et al., 2008).”

“BMI was included as a number of recent studies found that higher BMI is associated with greater brain atrophy in normal elderly subjects (Raji et al., 2009), and in MCI and AD (Ho et al., 2010b).”

“The hippocampal summaries were volumes generated from an automatic segmentation method that we developed based on machine learning; we recently validated this method against manual gold standards (Morra et al., 2008; Morra et al., 2009; Morra et al., 2010).”

“The hippocampal summaries were volumes generated from an automatic segmentation method that we developed based on machine learning; we recently validated this method against manual gold standards (Morra et al., 2008; Morra et al., 2009; Morra et al., 2010).”

“As we noted in Hua et al. (2010b), treatments may slow atrophy to different degrees, which may be denoted by k %, for different k .”

“PET, for example, offers metabolic or perfusion-based information that complements measures of structural atrophy on MRI (Fan et al., 2008; Hinrichs et al., 2009).”

“More sensitive and specific markers of early AD progression will help monitor the effectiveness of new treatments, and lessen the time and cost of clinical trials.”

“We also considered ApoE genotype (coded as 0, 1 or 2 for the number of E4 alleles), age, sex and BMI, as each can influence AD risk (Corder et al., 1993, Lindsay et al., 2002, Azad et al., 2007, Buchman et al., 2005).”

“When k is less than about 33%, the power estimates for AD subjects are improved compared with the minimal sample size of AD 48 subjects reported by Hua et al. (2009). There is a drop in the sample sizes needed to show a specific slowing effect, as the more AD-like subjects are selected.”

“It is worth noting that MRI-based machine learning has been used widely for classification not only for AD, but also for predicting changes in patients with brain tumors (Lukas et al., 2004), aphasia (Wilson et al.”

“The hippocampal summaries were volumes generated from an automatic segmentation method that we developed based on machine learning; we recently validated this method against manual gold standards (Morra et al., 2008; Morra et al., 2009; Morra et al., 2010).”

“MRI-based temporal and ventricular volumes have also been reported for their predictive power in MCI subjects (Fleisher et al., 2008; Korf et al., 2004).”

“A similar approach to boost the power of imaging biomarkers is voxel-set pre-selection, which substantially boosts power to detect the slowing of atrophy (Hua et al., 2010; Chen et al., 2010).”

“The few ADNI studies that have assessed the field strength effect (Ho et al., 2009; Kruggel et al., 2010) suggest that 1.”

“Unexpectedly, however, the MRI hippocampal summaries were not incorporated into our predictive model, which is surprising as hippocampal volume can be useful for prediction of MCI progression to AD (Apostolova et al., 2006a; Apostolova et al., 2006b; Apostolova et al., 2007; Frisoni et al., 2010).”

“We and others have found that a classifier can be given and entire brain image, and from it can derive the voxels whose signals are most promising for group classification (Sun et al., 2009).”

“Furthermore, machine learning can perhaps be used to discover genetic (Stein et al., 2009; Stein et al., 2010), epidemiological and physiological factors that influence the progression of AD.”

“…progression of the disease; in general, markers of amyloid deposition are thought to rise earlier than markers of neurodegeneration detectable on MRI, and these in turn become abnormal before tests of clinical function (Braskie et al., 2008; Jack et al., 2010; Petersen, 2010; Protas et al., 2010).”

“…mild; there is also substantial cross-subject variation in ventricular volume, even in healthy subjects (Chou et al., 2009b), and this may throw off a classifier’s accuracy unless the disease effect outweighs this natural variation (Chou et al., 2008;Chou et al., 2009a;Chou et al., 2009b).”

“…been used widely for classification not only for AD, but also for predicting changes in patients with brain tumors (Lukas et al., 2004), aphasia (Wilson et al., 2009), autism (Ecker et al., 2010), psychosis (Koutsouleris et al., 2009) and even for classifying patterns of brain activation in…”

“As we noted in Hua et al. (2010b), treatments may slow atrophy to different degrees, which may be denoted by k %, for different k .”

“BMI was included as a number of recent studies found that higher BMI is associated with greater brain atrophy in normal elderly subjects (Raji et al., 2009), and in MCI and AD (Ho et al., 2010b).”

“…mild; there is also substantial cross-subject variation in ventricular volume, even in healthy subjects (Chou et al., 2009b), and this may throw off a classifier’s accuracy unless the disease effect outweighs this natural variation (Chou et al., 2008;Chou et al., 2009a;Chou et al., 2009b).”

“By comparing different imaging measures (voxel-based, ROI-based, or surface-based; Gutman et al., 2008), and different classifiers (SVM versus others), future studies may be able to gauge which aspects of the classifier (its mathematical design or the features used) are most relevant for boosting…”

“…(Klunk et al., 2004; Protas et al., 2010), metabolic decline or perfusion deficits assessed by fluoro-deoxyglucose PET (PET-FDG), brain atrophy on MRI, and risk factors that influence these measures (e.g. ApoE, cardiovascular risks, etc.) (Frisoni et al., 2010; Jack et al., 2010; Petersen, 2010).”

“…mild; there is also substantial cross-subject variation in ventricular volume, even in healthy subjects (Chou et al., 2009b), and this may throw off a classifier’s accuracy unless the disease effect outweighs this natural variation (Chou et al., 2008;Chou et al., 2009a;Chou et al., 2009b).”

“Several studies have performed diagnostic classification based on MRI scans, using measures such as whole-brain patterns of atrophy (Davatzikos et al., 2009; Mesrob et al., 2008), tissue densities from voxel-based morphometry (Vemuri et al., 2008) and cortical thickness (Lerch et al., 2008).”

“This result supports the concept of clinical trial enrichment, which has been previously advocated (Cummings et al., 2007; Frisoni et al., 2010; Hampel and Broich, 2009).”

“…deposition, measured by Positron Emission Tomography (PET) and radiotracers that bind to the plaques and tangles in the brain (Klunk et al., 2004; Protas et al., 2010), metabolic decline or perfusion deficits assessed by fluoro-deoxyglucose PET (PET-FDG), brain atrophy on MRI, and risk factors…”

“Specifically, in conference abstracts, Alexander et al. (2008) and Zhang et al. (2008) have advocated a multivariate network analysis in which a very large number of regional brain volumes are jointly used as predictors, in an SVM model.”

“Specifically, in conference abstracts, Alexander et al. (2008) and Zhang et al. (2008) have advocated a multivariate network analysis in which a very large number of regional brain volumes are jointly used as predictors, in an SVM model.”

“Combined modalities The approach of Kohannim et al [86] combined multiple factors, including MRI and FDG-PET measures, CSF biomarkers, APOE genotype, age, sex, and BMI, to enhance machine learning methods for AD diagnosis.”

“As for the fusion ofmultiplemodalities includingMRI, PET, biological and neurological data for discriminating AD/MCI patients from healthy Normal Control (NC), Kohannim et al. (2010) concatenated features from modalities into a vector and used a Support Vector Machine (SVM) classifier.”

“Similarly, Kohannim et al. (2010) combined hippocampal, ventricular, and temporal lobe volumes, FDG-PET numeric summary, CSF biomarkers, APOE genotype, age, sex, and body mass index for SVM-based AD and MCI classification.”

“Kohannim et al. (2010) concatenated features from modalities into a vector and used a support vector machine (SVM) classifier.”