There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Jennifer Bramen |
| Institution: | Pacific Neuroscience Institute Foundation |
| Department: | Pacific Brain Health Center |
| Country: | |
| Proposed Analysis: | Statistical analysis plan (SAP) Summary for the development of a novel, sMRI-based biomarker to quantify disease progression in patients with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) using ADNI data: The focus of this study involves investigating anatomical features extracted from structural MRI in the context of different mathematical operations and machine learning algorithms to derive a biomarker for AD-related neurodegeneration that is optimally sensitive to disease severity, and predictive of diagnosis and progression. This is based upon the premise, evidenced by large bodies of previous research on patients with AD, that one would expect to observe regional brain volumes and cognitive metrics change in reflection of AD progression, based on a patterned underlying relationship. The five novel approaches to be investigated for deriving a single measure of AD-related neurodegeneration are distance metrics which take a list of normalized regional brain volumes significantly affected in AD neuropathology as inputs, and for each of the study groups (HC, MCI, and AD), compute their distance to a corresponding list of structural volumes extracted from the MNI305 template. These five distance-based biomarkers will include the Euclidean distance, and the Hausdorff and Fréchet distances in 1 and n dimensions, respectively. The Euclidean distance is calculated by treating the list of normalized regional volumes as a vector in ℝN and computing the Euclidean distance between this vector for each of the study groups, respectively, and that of the MNI305 template. The Hausdorff distance measures how far two subspaces of a metric space are from each other, while the Fréchet Distance quantifies the similarity between two curves or sets of points in space with an emphasis on the location and ordering of points. The Hausdorff and Fréchet distances are similar to one another, but the Hausdorff distance does not take into account the flow or order of points; it is the greatest of all distances from a point in one subset to the closest point in the other subset. The Hausdorff and Fréchet distances will each be calculated and assessed for efficacy as a biomarker in both 1 and n dimensions. The rationale for embedding these projected spaces in more dimensions of ℝN is elaborated on in Methods 2.1.4 of our full SAP, but in summary, is intended to increase the measurement accuracy for any given independent variable affected by many other variables, relying on the False Nearest Neighbors (FNN) algorithm to determine the correct number of embedding dimensions. An underlying assumption adopted here is that there are multiple variables that affect normalized regional volume; thus by determining the number of variables that affect this in a significant way and embedding the list of volumes in this space, the projected points are distributed spatially in ℝN in a way that reflects the information ascribed by each variable that affects normalized regional volume. The proposed study aims to evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict diagnosis, disease severity, and progression. The way in which these aims will be operationalized in the proposed study is outlined on the next page. Aims of the present study Aim 1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) diagnosis, operationalized as diagnosis at baseline, and (b) disease severity, operationalized as (I) the respective p-value for each metric in each pair-wise comparison possible for the 3 groups (Healthy Control - HC, MCI, and AD) and (II) Mini-Mental State Examination (MMSE) score at baseline Aim 1.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) disease state (HC, MCI, or AD) using a Gaussian Naïve Bayes (GNB) algorithm for the distance metrics, compared to said ability for the adjusted hippocampal volume (AHV) and MMSE score, also using the GNB algorithm, and (b) their sensitivity to disease severity (I) as measured by the respective p-values of each metric comparing patients with AD, MCI, and HC in every pair-wise combination of these 3 groups, and (II) their ability to correctly predict MMSE score class at baseline using a GNB algorithm for the distance metrics and AHV, as measured by classification performance. Aim 2: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, operationalized as (I) the respective p-value for the percent baseline change in each metric for each pair-wise comparison for the 3 study groups, (II) the relationship between the change in the biomarker to the change in MMSE score (from baseline to follow-up), and (III) the ability to predict conversion from one group to another using the change in the biomarker and baseline group classification as inputs to a GNB algorithm. Aim 2.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, as measured by (I) the respective p-values of the percent change in baseline for each metric for every pair-wise comparison possible for the 3 study groups, (II) the relationship between the change in each biomarker metric (from baseline to follow-up) to MMSE score, using linear regression and the mean-squared error minimization of a non-linear objective function, and (III) their ability to correctly predict conversion at follow-up using a GNB algorithm for the distance metrics and AHV. For a full SAP with the proposed study background, significance, hypotheses, and methods elaborately explained, please do not hesitate to contact us. |
| Additional Investigators | |
| Investigator's Name: | Emily Popa |
| Proposed Analysis: | Statistical analysis plan (SAP) Summary for the development of a novel, sMRI-based biomarker to quantify disease progression in patients with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) using ADNI data: The focus of this study involves investigating anatomical features extracted from structural MRI in the context of different mathematical operations and machine learning algorithms to derive a biomarker for AD-related neurodegeneration that is optimally sensitive to disease severity, and predictive of diagnosis and progression. This is based upon the premise, evidenced by large bodies of previous research on patients with AD, that one would expect to observe regional brain volumes and cognitive metrics change in reflection of AD progression, based on a patterned underlying relationship. The five novel approaches to be investigated for deriving a single measure of AD-related neurodegeneration are distance metrics which take a list of normalized regional brain volumes significantly affected in AD neuropathology as inputs, and for each of the study groups (HC, MCI, and AD), compute their distance to a corresponding list of structural volumes extracted from the MNI305 template. These five distance-based biomarkers will include the Euclidean distance, and the Hausdorff and Fréchet distances in 1 and n dimensions, respectively. The Euclidean distance is calculated by treating the list of normalized regional volumes as a vector in ℝN and computing the Euclidean distance between this vector for each of the study groups, respectively, and that of the MNI305 template. The Hausdorff distance measures how far two subspaces of a metric space are from each other, while the Fréchet Distance quantifies the similarity between two curves or sets of points in space with an emphasis on the location and ordering of points. The Hausdorff and Fréchet distances are similar to one another, but the Hausdorff distance does not take into account the flow or order of points; it is the greatest of all distances from a point in one subset to the closest point in the other subset. The Hausdorff and Fréchet distances will each be calculated and assessed for efficacy as a biomarker in both 1 and n dimensions. The rationale for embedding these projected spaces in more dimensions of ℝN is elaborated on in Methods 2.1.4 of our full SAP, but in summary, is intended to increase the measurement accuracy for any given independent variable affected by many other variables, relying on the False Nearest Neighbors (FNN) algorithm to determine the correct number of embedding dimensions. An underlying assumption adopted here is that there are multiple variables that affect normalized regional volume; thus by determining the number of variables that affect this in a significant way and embedding the list of volumes in this space, the projected points are distributed spatially in ℝN in a way that reflects the information ascribed by each variable that affects normalized regional volume. The proposed study aims to evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict diagnosis, disease severity, and progression. The way in which these aims will be operationalized in the proposed study is outlined on the next page. Aims of the present study Aim 1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) diagnosis, operationalized as diagnosis at baseline, and (b) disease severity, operationalized as (I) the respective p-value for each metric in each pair-wise comparison possible for the 3 groups (Healthy Control - HC, MCI, and AD) and (II) Mini-Mental State Examination (MMSE) score at baseline Aim 1.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) disease state (HC, MCI, or AD) using a Gaussian Naïve Bayes (GNB) algorithm for the distance metrics, compared to said ability for the adjusted hippocampal volume (AHV) and MMSE score, also using the GNB algorithm, and (b) their sensitivity to disease severity (I) as measured by the respective p-values of each metric comparing patients with AD, MCI, and HC in every pair-wise combination of these 3 groups, and (II) their ability to correctly predict MMSE score class at baseline using a GNB algorithm for the distance metrics and AHV, as measured by classification performance. Aim 2: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, operationalized as (I) the respective p-value for the percent baseline change in each metric for each pair-wise comparison for the 3 study groups, (II) the relationship between the change in the biomarker to the change in MMSE score (from baseline to follow-up), and (III) the ability to predict conversion from one group to another using the change in the biomarker and baseline group classification as inputs to a GNB algorithm. Aim 2.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, as measured by (I) the respective p-values of the percent change in baseline for each metric for every pair-wise comparison possible for the 3 study groups, (II) the relationship between the change in each biomarker metric (from baseline to follow-up) to MMSE score, using linear regression and the mean-squared error minimization of a non-linear objective function, and (III) their ability to correctly predict conversion at follow-up using a GNB algorithm for the distance metrics and AHV. For a full SAP with the proposed study background, significance, hypotheses, and methods elaborately explained, please do not hesitate to contact us. |
| Investigator's Name: | Paul Thompson |
| Proposed Analysis: | Statistical analysis plan (SAP) Summary for the development of a novel, sMRI-based biomarker to quantify disease progression in patients with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) using ADNI data: The focus of this study involves investigating anatomical features extracted from structural MRI in the context of different mathematical operations and machine learning algorithms to derive a biomarker for AD-related neurodegeneration that is optimally sensitive to disease severity, and predictive of diagnosis and progression. This is based upon the premise, evidenced by large bodies of previous research on patients with AD, that one would expect to observe regional brain volumes and cognitive metrics change in reflection of AD progression, based on a patterned underlying relationship. The five novel approaches to be investigated for deriving a single measure of AD-related neurodegeneration are distance metrics which take a list of normalized regional brain volumes significantly affected in AD neuropathology as inputs, and for each of the study groups (HC, MCI, and AD), compute their distance to a corresponding list of structural volumes extracted from the MNI305 template. These five distance-based biomarkers will include the Euclidean distance, and the Hausdorff and Fréchet distances in 1 and n dimensions, respectively. The Euclidean distance is calculated by treating the list of normalized regional volumes as a vector in ℝN and computing the Euclidean distance between this vector for each of the study groups, respectively, and that of the MNI305 template. The Hausdorff distance measures how far two subspaces of a metric space are from each other, while the Fréchet Distance quantifies the similarity between two curves or sets of points in space with an emphasis on the location and ordering of points. The Hausdorff and Fréchet distances are similar to one another, but the Hausdorff distance does not take into account the flow or order of points; it is the greatest of all distances from a point in one subset to the closest point in the other subset. The Hausdorff and Fréchet distances will each be calculated and assessed for efficacy as a biomarker in both 1 and n dimensions. The rationale for embedding these projected spaces in more dimensions of ℝN is elaborated on in Methods 2.1.4 of our full SAP, but in summary, is intended to increase the measurement accuracy for any given independent variable affected by many other variables, relying on the False Nearest Neighbors (FNN) algorithm to determine the correct number of embedding dimensions. An underlying assumption adopted here is that there are multiple variables that affect normalized regional volume; thus by determining the number of variables that affect this in a significant way and embedding the list of volumes in this space, the projected points are distributed spatially in ℝN in a way that reflects the information ascribed by each variable that affects normalized regional volume. The proposed study aims to evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict diagnosis, disease severity, and progression. The way in which these aims will be operationalized in the proposed study is outlined on the next page. Aims of the present study Aim 1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) diagnosis, operationalized as diagnosis at baseline, and (b) disease severity, operationalized as (I) the respective p-value for each metric in each pair-wise comparison possible for the 3 groups (Healthy Control - HC, MCI, and AD) and (II) Mini-Mental State Examination (MMSE) score at baseline Aim 1.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) disease state (HC, MCI, or AD) using a Gaussian Naïve Bayes (GNB) algorithm for the distance metrics, compared to said ability for the adjusted hippocampal volume (AHV) and MMSE score, also using the GNB algorithm, and (b) their sensitivity to disease severity (I) as measured by the respective p-values of each metric comparing patients with AD, MCI, and HC in every pair-wise combination of these 3 groups, and (II) their ability to correctly predict MMSE score class at baseline using a GNB algorithm for the distance metrics and AHV, as measured by classification performance. Aim 2: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, operationalized as (I) the respective p-value for the percent baseline change in each metric for each pair-wise comparison for the 3 study groups, (II) the relationship between the change in the biomarker to the change in MMSE score (from baseline to follow-up), and (III) the ability to predict conversion from one group to another using the change in the biomarker and baseline group classification as inputs to a GNB algorithm. Aim 2.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, as measured by (I) the respective p-values of the percent change in baseline for each metric for every pair-wise comparison possible for the 3 study groups, (II) the relationship between the change in each biomarker metric (from baseline to follow-up) to MMSE score, using linear regression and the mean-squared error minimization of a non-linear objective function, and (III) their ability to correctly predict conversion at follow-up using a GNB algorithm for the distance metrics and AHV. For a full SAP with the proposed study background, significance, hypotheses, and methods elaborately explained, please do not hesitate to contact us. |
| Investigator's Name: | Susan Bookheimer |
| Proposed Analysis: | Statistical analysis plan (SAP) Summary for the development of a novel, sMRI-based biomarker to quantify disease progression in patients with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) using ADNI data: The focus of this study involves investigating anatomical features extracted from structural MRI in the context of different mathematical operations and machine learning algorithms to derive a biomarker for AD-related neurodegeneration that is optimally sensitive to disease severity, and predictive of diagnosis and progression. This is based upon the premise, evidenced by large bodies of previous research on patients with AD, that one would expect to observe regional brain volumes and cognitive metrics change in reflection of AD progression, based on a patterned underlying relationship. The five novel approaches to be investigated for deriving a single measure of AD-related neurodegeneration are distance metrics which take a list of normalized regional brain volumes significantly affected in AD neuropathology as inputs, and for each of the study groups (HC, MCI, and AD), compute their distance to a corresponding list of structural volumes extracted from the MNI305 template. These five distance-based biomarkers will include the Euclidean distance, and the Hausdorff and Fréchet distances in 1 and n dimensions, respectively. The Euclidean distance is calculated by treating the list of normalized regional volumes as a vector in ℝN and computing the Euclidean distance between this vector for each of the study groups, respectively, and that of the MNI305 template. The Hausdorff distance measures how far two subspaces of a metric space are from each other, while the Fréchet Distance quantifies the similarity between two curves or sets of points in space with an emphasis on the location and ordering of points. The Hausdorff and Fréchet distances are similar to one another, but the Hausdorff distance does not take into account the flow or order of points; it is the greatest of all distances from a point in one subset to the closest point in the other subset. The Hausdorff and Fréchet distances will each be calculated and assessed for efficacy as a biomarker in both 1 and n dimensions. The rationale for embedding these projected spaces in more dimensions of ℝN is elaborated on in Methods 2.1.4 of our full SAP, but in summary, is intended to increase the measurement accuracy for any given independent variable affected by many other variables, relying on the False Nearest Neighbors (FNN) algorithm to determine the correct number of embedding dimensions. An underlying assumption adopted here is that there are multiple variables that affect normalized regional volume; thus by determining the number of variables that affect this in a significant way and embedding the list of volumes in this space, the projected points are distributed spatially in ℝN in a way that reflects the information ascribed by each variable that affects normalized regional volume. The proposed study aims to evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict diagnosis, disease severity, and progression. The way in which these aims will be operationalized in the proposed study is outlined on the next page. Aims of the present study Aim 1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) diagnosis, operationalized as diagnosis at baseline, and (b) disease severity, operationalized as (I) the respective p-value for each metric in each pair-wise comparison possible for the 3 groups (Healthy Control - HC, MCI, and AD) and (II) Mini-Mental State Examination (MMSE) score at baseline Aim 1.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) disease state (HC, MCI, or AD) using a Gaussian Naïve Bayes (GNB) algorithm for the distance metrics, compared to said ability for the adjusted hippocampal volume (AHV) and MMSE score, also using the GNB algorithm, and (b) their sensitivity to disease severity (I) as measured by the respective p-values of each metric comparing patients with AD, MCI, and HC in every pair-wise combination of these 3 groups, and (II) their ability to correctly predict MMSE score class at baseline using a GNB algorithm for the distance metrics and AHV, as measured by classification performance. Aim 2: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, operationalized as (I) the respective p-value for the percent baseline change in each metric for each pair-wise comparison for the 3 study groups, (II) the relationship between the change in the biomarker to the change in MMSE score (from baseline to follow-up), and (III) the ability to predict conversion from one group to another using the change in the biomarker and baseline group classification as inputs to a GNB algorithm. Aim 2.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, as measured by (I) the respective p-values of the percent change in baseline for each metric for every pair-wise comparison possible for the 3 study groups, (II) the relationship between the change in each biomarker metric (from baseline to follow-up) to MMSE score, using linear regression and the mean-squared error minimization of a non-linear objective function, and (III) their ability to correctly predict conversion at follow-up using a GNB algorithm for the distance metrics and AHV. For a full SAP with the proposed study background, significance, hypotheses, and methods elaborately explained, please do not hesitate to contact us. |
| Investigator's Name: | Gavin Kress |
| Proposed Analysis: | Statistical analysis plan (SAP) Summary for the development of a novel, sMRI-based biomarker to quantify disease progression in patients with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) using ADNI data: The focus of this study involves investigating anatomical features extracted from structural MRI in the context of different mathematical operations and machine learning algorithms to derive a biomarker for AD-related neurodegeneration that is optimally sensitive to disease severity, and predictive of diagnosis and progression. This is based upon the premise, evidenced by large bodies of previous research on patients with AD, that one would expect to observe regional brain volumes and cognitive metrics change in reflection of AD progression, based on a patterned underlying relationship. The five novel approaches to be investigated for deriving a single measure of AD-related neurodegeneration are distance metrics which take a list of normalized regional brain volumes significantly affected in AD neuropathology as inputs, and for each of the study groups (HC, MCI, and AD), compute their distance to a corresponding list of structural volumes extracted from the MNI305 template. These five distance-based biomarkers will include the Euclidean distance, and the Hausdorff and Fréchet distances in 1 and n dimensions, respectively. The Euclidean distance is calculated by treating the list of normalized regional volumes as a vector in ℝN and computing the Euclidean distance between this vector for each of the study groups, respectively, and that of the MNI305 template. The Hausdorff distance measures how far two subspaces of a metric space are from each other, while the Fréchet Distance quantifies the similarity between two curves or sets of points in space with an emphasis on the location and ordering of points. The Hausdorff and Fréchet distances are similar to one another, but the Hausdorff distance does not take into account the flow or order of points; it is the greatest of all distances from a point in one subset to the closest point in the other subset. The Hausdorff and Fréchet distances will each be calculated and assessed for efficacy as a biomarker in both 1 and n dimensions. The rationale for embedding these projected spaces in more dimensions of ℝN is elaborated on in Methods 2.1.4 of our full SAP, but in summary, is intended to increase the measurement accuracy for any given independent variable affected by many other variables, relying on the False Nearest Neighbors (FNN) algorithm to determine the correct number of embedding dimensions. An underlying assumption adopted here is that there are multiple variables that affect normalized regional volume; thus by determining the number of variables that affect this in a significant way and embedding the list of volumes in this space, the projected points are distributed spatially in ℝN in a way that reflects the information ascribed by each variable that affects normalized regional volume. The proposed study aims to evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict diagnosis, disease severity, and progression. The way in which these aims will be operationalized in the proposed study is outlined on the next page. Aims of the present study Aim 1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) diagnosis, operationalized as diagnosis at baseline, and (b) disease severity, operationalized as (I) the respective p-value for each metric in each pair-wise comparison possible for the 3 groups (Healthy Control - HC, MCI, and AD) and (II) Mini-Mental State Examination (MMSE) score at baseline Aim 1.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict (a) disease state (HC, MCI, or AD) using a Gaussian Naïve Bayes (GNB) algorithm for the distance metrics, compared to said ability for the adjusted hippocampal volume (AHV) and MMSE score, also using the GNB algorithm, and (b) their sensitivity to disease severity (I) as measured by the respective p-values of each metric comparing patients with AD, MCI, and HC in every pair-wise combination of these 3 groups, and (II) their ability to correctly predict MMSE score class at baseline using a GNB algorithm for the distance metrics and AHV, as measured by classification performance. Aim 2: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, operationalized as (I) the respective p-value for the percent baseline change in each metric for each pair-wise comparison for the 3 study groups, (II) the relationship between the change in the biomarker to the change in MMSE score (from baseline to follow-up), and (III) the ability to predict conversion from one group to another using the change in the biomarker and baseline group classification as inputs to a GNB algorithm. Aim 2.1: Evaluate the efficacy of the Euclidean Distance, the Hausdorff Distance, calculated with and without embedding in higher dimensions, and the Fréchet Distance, also calculated with and without embedding in higher dimensions, as structural biomarkers of AD-related neurodegeneration on the basis of their ability to predict disease progression, as measured by (I) the respective p-values of the percent change in baseline for each metric for every pair-wise comparison possible for the 3 study groups, (II) the relationship between the change in each biomarker metric (from baseline to follow-up) to MMSE score, using linear regression and the mean-squared error minimization of a non-linear objective function, and (III) their ability to correctly predict conversion at follow-up using a GNB algorithm for the distance metrics and AHV. For a full SAP with the proposed study background, significance, hypotheses, and methods elaborately explained, please do not hesitate to contact us. |
