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).”

“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.”

“[90] investigated the usage of a support vector machine to boost the power of clinical trials and reduce the clinical trial sample size.”

“With this method, FDG PET had similar power to detect change as MRI methods [12].”

“Kohannim et al [86] investigated the utility of their machine learning classifier, based B) MMSE scores.”

“They are also useful for diagnosis and prognosis (Frisoni et al. 2010, Kohannim et al. 2010).”

“In Kohannim et al. [17], the improvement from using multiple biomarkers was not significant and Davatzikos et al. [15] reported marginal improvements which, however, may be related to the fact that results with only one biomarker were not very good to begin with.”

“…(SVMs; Cortes & Vapnik, 1995), for example, use machine learning to analyze, classify, and recognize patterns in data and have recently been tested in the classification of several diseases including Parkinson’s disease (Gil & Johnson, 2009) and Alzheimer’s disease ( Kohannim et al., 2010).”

“Kohannim et al. (2010) have also combined MRI, CSF and FDG-PET in a smaller cohort.”

“…cropping, denoising, resampling, intensity normalization, regions-of-interest (ROI) extraction (Cu-ingnet et al., 2011; Davatzikos et al., 2011; Kohannim et al., 2010), and feature selection (Liu et al., 2012, 2014a; Akha-van Aghdam et al., 2018; Qian et al., 2020), prepare the data for more…”

“For example, in Kohannim et al. (2010), a support vector machine (SVM) classifier was applied to the concatenated features from multiple modalities for the classification of AD.”

“This strategy increases the proportion of patients who could benefit from the intervention and optimizes the statistical power of the trial (Kohannim et al., 2010; Frisoni et al., 2010).”

“With respect to multi-modal fusion, it is known that different modalities can provide complementary information, and thus can enhance the diagnostic accuracy (Cui et al., 2011; Hinrichs et al., 2011; Kohannim et al., 2010; Perrin et al., 2009; Suk and Shen, 2013; Walhovd et al., 2010; Westman et al., 2012).”

“[36] Kohannim O, Hua X, Hibar DP, Lee S, Chou YY, Toga AW, et al.”

“Support vector machines (SVM) [15, 16] are the most commonly used algorithm in AD research for multivariate classification [9, 17–39].”

“…prior works on the automated disease diagnosis problem, that include partial least square based feature selection on MRI (Zhou et al., 2011), feature extraction methods based onMRI data (Cuingnet et al., 2011), and support vector machines to combine MRI, PET and CSF, etc. (Kohannim et al., 2010).”

“Targeting subpopulations most likely to show decline also boosts power for prevention and treatment trials (Kohannim et al., 2010).”

“In particular, Kohannim et al. (2010) combined multiple biomarkers, including MRI and FDG-PET measures as well as CSF and other biomarkers for disease status classification using SVM classifiers and reported an increase in power to predict future decline.”

“…the detection of AD and the prediction of subsequent clinical decline in MCI patients (Beckett et al., 2010; Chupin et al., 2009; Hua et al., 2008; Kohannim et al., 2010; Landau et al., 2010; Leung et al., 2010; Lorenzi et al., 2010; Morra et al., 2008; Risacher et al., 2009; Schott et al., 2010;…”

“Recent studies have shown that information fusion of multiple modalities can help enhance the diagnostic performance (Perrin et al. 2009; Kohannim et al. 2010; Walhovd et al. 2010; Cui et al. 2011; Hinrichs et al. 2011; Zhang et al. 2011; Westman et al. 2012; Yuan et al. 2012; Zhang and Shen 2012; Suk et al. 2015).”

“We hypothesized that our proposed multivariate morphometry feature may boost power by reducing sample size estimates required to observe effects, in both morphometry studies and potentially also in clinical trials (Hua et al., 2010; Kohannim et al., 2010).”

“ML tools have found their way into the medical fi eld for a number of different applications, ranging from health care services (24) to clinical predictions of the progression of Alzheimer ’ s disease (25,26).”

“Current approaches used to classify Alzheimer’s disease (Kohannim et al., 2010; Klöppel et al., 2008) rely on features such as volumetric measures from anatomical regions in magnetic resonance imaging (MRI) images of the brain, CSF biomarkers, ApoE genotype, age, sex, body mass index, and, in some…”

“With multivariate features, it is natural to apply machine learning methods to perform computer-assisted diagnosis and predict future clinical decline (Kohannim et al., 2010; Sun et al., 2009a;Wang et al., 2010a).”

“demographic data, structural and functional neuroimaging, genotyping information and laboratory results, such as cerebrospinal fluid biomarkers) in order to improve successful classification.(61)”

“A straightforward way to fusion multimodality data is to pool features from multiple modalities together [33,34], and then feed them to train a classifier.”

“• Conducting clinical trials using ML algorithms as a guidance for optimal treatment strategies.(45)”

“A variety of studies have been proposed to characterize and predict AD [5], [7]–[13] with the most important biomarkers detected as brain volume and thickness.”

“…for neuroimaging outcome measures make this assumption implicitly [Beckett et al., 2010; Cummings, 2010; Ho et al., 2010; Hua et al., 2009, 2010; Kohannim et al., 2010; Lorenzi et al., 2010; Nestor et al., 2008; Risacher et al., 2010; Schuff et al., 2009; Vemuri et al., 2010], with a few…”

“The identification of MCI as a possible prodrome to AD, as well as the recent development classifier algorithms that assess later risk for AD based on a variety of clinical factors [10], leaves open the potential for initiating therapies, including HRT, that may prevent progression to AD in those at risk.”

“Kohannim et al. [26] verified that directly concatenating multi-modal (MRI, PET, and cerebrospinal fluid) features results in better performance than using solely single-modal features.”

“In addition, there is reference to sKL-derived power estimates in Kohannim et al. (2010) and Ho et al.”

“HC subjects [113] or an MCI subject who converted to AD dementia versus MCI subjects remaining stable [114].”

“…publications using tensor based morphometry (TBM) measures of longitudinal change estimated from structural MRI available through the Alzheimer's Disease Neuroimaging Initiative (ADNI): Hua et al. (2008a,b, 2009, 2010); Ho et al. (2010); Kohannim et al. (2010); Beckett et al. (2010); among others.”

“combined brain imaging and other biomarkers to classify ADNI subjects as AD, MCI and NC (27).”

“, 2011), single photon emitting computer tomography (SPECT) (Fung and Stoeckel, 2007) or a combination of multi-source datasets (Calhoun and Adali, 2009; Chen et al., 2009; Correa et al., 2010; Groves et al., 2011; Jack et al., 2010; Kohannim et al., 2010; Sui et al., 2011; Yang et al., 2010; Yuan et al., 2012a).”

“[8] used the support vector machine (SVM) as a classifier on the concatenated features obtained from multiple biomarkers.”

“Kohannim et al. [3] concatenated features from multiple modalities including MRI, PET, biological and neurological data into a vector and used a Support Vector Machine (SVM).”

“In addition, the large number of statistical tests involved in a genome-wide and/or image wide search requires special methods to boost power, including gene-based tests [126], ridge regression models [127], multilocus modeling, and meta-analysis.”

“A concatenation of features from multiple modalities including MRI, PET, biological and neurological data into a vector (Kohannim et al. 2010) used an SVM as classifier.”

“(Kohannim et al., 2010) CN 213, AD 158 Referral sample ADNI Multi voxel classifier AUROC 0.”

“That is, hippocampal volume has been used extensively in previous AD imaging studies [14,16,25,26], and including it ensures continuity with this literature.”

“Subject selection by multiple biomarker classifier ,40 (,40) [86]”

“Using AD biomarkers as outcome measures in AD trials can similarly improve trial efficiency [9,27,28,31,35,57,59].”

“In some studies, the nuisance features are given as ordinary variables to the classifier, such as support vector machine, and it is assumed that the classifier is able to correct the interactions between nuisance features and classification features [13].”

“Many groups including ours have made important strides towards multimodal biomarker diagnostic discrimination (Cui et al., 2011; Davatzikos et al., 2011; Ewers et al., 2012; Kohannim et al., 2010; Vos et al., 2012; Walhovd et al., 2010; Westman et al., 2012).”

“1.”

“For example, Kohannim et al. (2010) concatenated the features extracted from various modalities into a vector and classified it using support vector machine (SVM) (Kohannim et al., 2010).”

“(1) For signal stabilization and to allow the participants to adapt to the environment, the first 10 EPI volumes of the fMRI images were discarded.”

“This 71–73 Enriching cohorts with subjects who are at the correct stage for each individual compound can be achieved by fundamentally categorizing based upon biomarkers that relate to homogenous staging and prospective endpoint declines.”

“Although discrepant findings exist, several studies show that MRI biomarkers are more accurate predictors of cognitive function and clinical change in MCI than CSF measures of amyloid [74, 76, 78–81].”

“There are other recent studies that have developed new methodologies to improve prediction accuracy based on machine learning tools combining information from FDG-PET, sMRI and other biomarkers [83, 89-95].”

“[156] selected a SVM model and performed leave-oneout CV for AD and MCI classification and prediction of future cognitive decline within 1 year, and achieved a maximum of 90% accuracy for AD vs healthy controls.”

“The combination with one or preferably several markers that reflects brain dysfunction or neurodegeneration has been suggested [33,34], such as with regional brain atrophy [35,36], regional cerebral hypoperfusion [37,38], decreased regional glucose metabolism [36,39], or altered cortical connectivity (through resting…”

“Besides structural MRI, other imaging modalities such as functional MRI can also be used in AD/MCI diagnosis 19–23 , as they provide additional functional information about hypometabolism and specific protein quantification, which can be beneficial in disease diagnosis.”

“zp is the value of the standard Gaussian distribution for P[Z < zp] = p [8].”

“A variety of measures have proven capable of distinguishing MCI from cognitively normal individuals, including clinical evaluation [9,10], neuroimaging [9,11–19], and biochemical [16,19] measures.”

“al [4] combine FDG-PET-derived numerical summaries, MRI-derived volume measures, CSF biomarkers, APOE genotype, and subject demographics for the task of discriminating MCI from AD.”

“In this paper, we denote matrices as boldface uppercase letters, vectors as boldface lowercase letters, and scalars as normal italic letters, respectively.”

“For example, support vector machines [40] is used for the analysis, classification, and recognition of Parkinson’s [41] and Alzheimer’s diseases [42].”

“…the combined role of different biomarkers such as CSF and MRI markers (Davatzikos et al., 2011; Westman et al., 2012), 18 F-FDG and 11C-PIB-PET (Zhang et al., 2012), MRI measures, PET-FDG numerical summary, CSF biomarkers (t-tau, p-tau, and A β 42), and APOE genotype (Kohannim et al., 2010).”

“healthy elderly controls [78].”

“Although many studies have used machine learning to predict the progression of neurodegenerative diseases and differentiate diagnostic groups such as AD, MCI, and controls (Kloppel et al., 2012; Kohannim et al., 2010; Vemuri et al., 2008), we found no attempts in the literature that used learning to directly optimize…”

“Multivariate methods can also combine multiple biomarkers to select subjects more likely to show detectable change in hypothetical AD trials (Kohannim et al., 2010). served.”

“[48] directly concatenated features of multimodal data into a long vector and then used SVM for AD classification.”