MSmcDESPOT: Follow-Ups November 1, 2010. Where We Are Baseline cross-section conclusions: – DVF is sensitive to early stages of MS where other measures.

Slides:



Advertisements
Similar presentations
Corpus Callosum Damage Predicts Disability Progression and Cognitive Dysfunction in Primary-Progressive MS After Five Years.
Advertisements

A small taste of inferential statistics
Introduction and Aim Multiple sclerosis (MS) is a chronic neurological disease involving demyelination of the nervous system. There are three key MS sub-types:
CHAPTER 21 Inferential Statistical Analysis. Understanding probability The idea of probability is central to inferential statistics. It means the chance.
Linear Regression t-Tests Cardiovascular fitness among skiers.
Table of Contents Exit Appendix Behavioral Statistics.
Introduction to Statistical Quality Control, 4th Edition Chapter 7 Process and Measurement System Capability Analysis.
ANOVA: PART II. Last week  Introduced to a new test:  One-Way ANOVA  ANOVA’s are used to minimize family-wise error:  If the ANOVA is statistically.
Some Terms Y =  o +  1 X Regression of Y on X Regress Y on X X called independent variable or predictor variable or covariate or factor Which factors.
MSmcDESPOT A look at the road behind and ahead October 30, 2009.
ISMRM 2010 Quantitative Imaging and MS. N. D. Gai and J. A. Butman, NIH T1 Error Analysis for Double Angle Technique and Comparison to Inversion Recovery.
Introduction: The lesion-centered view on MS __________________________________________________________________ RRI/TUD/StanU – HH Kitzler Specific Aims:
Declaration of Conflict of Interest or Relationship I have no conflicts of interest to disclose with regard to the subject matter of this presentation.
The Basics of Regression continued
An Overview of Today’s Class
Lecture 9: One Way ANOVA Between Subjects
1 Psych 5500/6500 The t Test for a Single Group Mean (Part 5): Outliers Fall, 2008.
MSmcDESPOT: Baseline vs. 1- year Diagnosis. N008 Baseline SPGR.
REGRESSION AND CORRELATION
Dealing with Heteroscedasticity In some cases an appropriate scaling of the data is the best way to deal with heteroscedasticity. For example, in the model.
1 BA 555 Practical Business Analysis Review of Statistics Confidence Interval Estimation Hypothesis Testing Linear Regression Analysis Introduction Case.
Stat 112: Lecture 9 Notes Homework 3: Due next Thursday
Correlation and Regression Analysis
Linear Regression 2 Sociology 5811 Lecture 21 Copyright © 2005 by Evan Schofer Do not copy or distribute without permission.
So are how the computer determines the size of the intercept and the slope respectively in an OLS regression The OLS equations give a nice, clear intuitive.
Psy B07 Chapter 1Slide 1 ANALYSIS OF VARIANCE. Psy B07 Chapter 1Slide 2 t-test refresher  In chapter 7 we talked about analyses that could be conducted.
Composite MRI scores improve correlation with EDSS in multiple sclerosis by Poonawalla et al. Review by Jason Su.
1 Sampling Distributions Presentation 2 Sampling Distribution of sample proportions Sampling Distribution of sample means.
Introduction to Statistical Quality Control, 4th Edition Chapter 7 Process and Measurement System Capability Analysis.
Relationships Scatterplots and correlation BPS chapter 4 © 2006 W.H. Freeman and Company.
Inference for Regression
Inferences for Regression
Bootstrap and Cross-Validation Bootstrap and Cross-Validation.
F OUNDATIONS OF S TATISTICAL I NFERENCE. D EFINITIONS Statistical inference is the process of reaching conclusions about characteristics of an entire.
Slides to accompany Weathington, Cunningham & Pittenger (2010), Chapter 3: The Foundations of Research 1.
7. Comparing Two Groups Goal: Use CI and/or significance test to compare means (quantitative variable) proportions (categorical variable) Group 1 Group.
MGS3100_04.ppt/Sep 29, 2015/Page 1 Georgia State University - Confidential MGS 3100 Business Analysis Regression Sep 29 and 30, 2015.
Sampling Error.  When we take a sample, our results will not exactly equal the correct results for the whole population. That is, our results will be.
Chapter 8 Delving Into The Use of Inference 8.1 Estimating with Confidence 8.2 Use and Abuse of Tests.
Educational Research Chapter 13 Inferential Statistics Gay, Mills, and Airasian 10 th Edition.
MSmcDESPOT A Brief Summary April 2, The Technique mcDESPOT (multi-component driven equilibrium single pulse observation of T1/T2) is a quantitative.
Testing hypotheses Continuous variables. H H H H H L H L L L L L H H L H L H H L High Murder Low Murder Low Income 31 High Income 24 High Murder Low Murder.
CROSS-VALIDATION AND MODEL SELECTION Many Slides are from: Dr. Thomas Jensen -Expedia.com and Prof. Olga Veksler - CS Learning and Computer Vision.
Copyright © 2013, 2009, and 2007, Pearson Education, Inc. Chapter 14 Comparing Groups: Analysis of Variance Methods Section 14.1 One-Way ANOVA: Comparing.
Stat 112 Notes 9 Today: –Multicollinearity (Chapter 4.6) –Multiple regression and causal inference.
Statistics in Applied Science and Technology Chapter14. Nonparametric Methods.
Relationships Scatterplots and Correlation.  Explanatory and response variables  Displaying relationships: scatterplots  Interpreting scatterplots.
Week 6. Statistics etc. GRS LX 865 Topics in Linguistics.
Other Types of t-tests Recapitulation Recapitulation 1. Still dealing with random samples. 2. However, they are partitioned into two subsamples. 3. Interest.
Displaying your data and using Classify Exploring how to use the legend classify command.
Unit 1: Representing Data & Analysing 2D Data 1.2 Understanding Variability in Data.
Sampling Distributions & Sample Means Movie Clip.
MPRAGEpre – Image Quality Quality is fairly consistent throughout subjects but there are a couple notable outliers: P003 & P025.
Chapter 13 Understanding research results: statistical inference.
P025 MPRAGE Pre-Contrast. P025 MPRAGE w/ Z-Score < -4.
King Faisal University جامعة الملك فيصل Deanship of E-Learning and Distance Education عمادة التعلم الإلكتروني والتعليم عن بعد [ ] 1 جامعة الملك فيصل عمادة.
Statistics 19 Confidence Intervals for Proportions.
CHAPTER 11 Mean and Standard Deviation. BOX AND WHISKER PLOTS  Worksheet on Interpreting and making a box and whisker plot in the calculator.
Statistical Concepts Basic Principles An Overview of Today’s Class What: Inductive inference on characterizing a population Why : How will doing this allow.
Choosing and using your statistic. Steps of hypothesis testing 1. Establish the null hypothesis, H 0. 2.Establish the alternate hypothesis: H 1. 3.Decide.
Baseline and longitudinal patterns of brain atrophy in MCI patients, and their use in prediction of Short- term conversion to AD: Results from ADNI Xuejiao.
Dependent-Samples t-Test
Statistics in MSmcDESPOT
ISMRM 2011 E-Poster #4643 mcDESPOT-Derived MWF Improves EDSS Prediction in MS Patients Compared to Atrophy Measures Alone J.Su1, H.H.Kitzler2, M.Zeineh1,
CHAPTER 29: Multiple Regression*
Gerald Dyer, Jr., MPH October 20, 2016
Brain Fraction and SPM Normalized Penumbra
Process and Measurement System Capability Analysis
ISMRM 2012 Prelim. Abstracts Oct 17, 2011 – Jason Su
Professor John Canny Fall 2004
Presentation transcript:

MSmcDESPOT: Follow-Ups November 1, 2010

Where We Are Baseline cross-section conclusions: – DVF is sensitive to early stages of MS where other measures are not – DVF correlates with EDSS (R^2 = 0.37 in NAWM) – The addition of a quantitative measure significantly improves EDSS prediction compared to volumetric atrophy measures alone 1yr follow-up scans – 23/26 patients scanned – 5/26 normals scanned and more incoming

Patient Overview

Correlation Plots

Discussion Our correlations for DV and DVF were consistent between the baseline and follow up Change in DV has a large scale – Could mean we’re quite sensitive to changes in the brain – Need to quantify how much DV varies due to repeatability error DV increased for almost every patient, only P020 had a drop (not shown because no EDSS data) Currently, we unexpectedly observe a negative correlation between change in EDSS and change in DV

Normals N004, N008, N012 First glimpse at the repeatability of the DV measure and mcDESPOT-derived MWF maps We would like to see little change between the baseline and follow up MWF maps Mean MWF BaselineStd. Dev.Mean MWF 1yrStd. Dev. N N N

Correlations

Histograms

DV Baseline DV (mm^3) 1yr DV (mm^3) N N N P Using the old baseline mean and std. dev. MWF maps, computed DV for the new normal scans Disconcertingly large increase in DV Why? – Biased using the baseline mean derived from normals to get their baseline DV – Follow-up scan quality?

Baseline SPGR_fa13 Images

1yr SPGR_fa13 Images

Discussion I would argue that the reduction in quality in the follow-up scans is comparable between the normals and P022 Are there ways to deal with the bias? – Cross-validation Try many random subsets of the normal population to generate mean and std. dev., choose the map pair that minimizes total DV among all normals – Ensemble methods Use all the map pairs and for each, generate a DV mask, then a voxel is considered demyelinated only if a majority of the DV masks have it as demyelinated

MSmcDESPOT: Looking at Maps October 29, 2010

Motivation Thus far we’ve been studying DV and DVF, which collapses all of our data into a single metric for each patient One of the key advantages of mcDESPOT is that it acquires whole brain maps We should start looking at our data as whole brain maps – Perhaps different subtypes of MS are associated with different spatial distributions of MWF

Baseline: Mean MWF Normals

Baseline: Mean MWF CIS

Baseline: Mean MWF RRMS

Baseline: Mean MWF SPMS

Baseline: Mean MWF PPMS

Discussion There’s clearly a drop in overall MWF as we progress from CIS to RR to SP to PP Can’t really discern any favoring for locations of low MWF other than around the ventricles – DV maps would probably show this better than anything, should generate a probabilistic DV map

Baseline: Std. Dev. MWF Normals

Baseline: Std. Dev. MWF CIS

Baseline: Std. Dev. MWF RRMS

Baseline: Std. Dev. MWF SPMS

Baseline: Std. Dev. MWF PPMS

Discussion In normals, MWF has a much lower standard deviation in WM areas RR patients seem to have an overall lower standard deviation than CIS – One interpretation might be that CIS patients are only starting to lose myelin so there is a lot of variability among them PP is by far the worst, the variance of MWF among the subjects seems to be the same throughout the brain – This means that the amount and location of myelin lost among PP patients varies wildly Of course standard deviation is a group based measure, not sure about the direct clinical application for one patient The 1yr cross-section maps looks like the baseline

Difference Maps For each subject, the difference map was computed as MWF_1yr – MWF_baseline – Then the mean difference between patients was computed for each subtype as well as the standard deviation of the differences The following maps may be hard to look at, they are highly non-traditional and probably it’s the first time anyone has ever seen such images

Difference: Mean CIS

Difference: Mean RRMS

Difference: Mean SPMS

Discussion There is a clear different between CIS and RR, with RR patients having much larger drops in MWF Actually, I feel like RR patients have the most actively changing MWF among all the subtypes looking at these images – Consistent with early stages being the most active? Have to check the ages of our RR patients.

Ratio Maps For each subject, the ratio map was computed as MWF_1yr/MWF_baseline – Then the mean ratio between patients was computed for each subtype as well as the standard deviation of the ratios These maps are ugly, it is tough to tell what’s going on – Ignore the white fringing around the brain, caused by regions of low MWF – Inside the brain, they would indicate places where lesions with low MWF are Maybe even they show lesions that have remyelinated a little as (not as small MWF)/(really small MWF) = big number

Ratio: Mean CIS

Ratio: Mean RRMS

Ratio: Mean SPMS

Ratio: Mean PPMS

Discussion Hard to decipher these – CIS seems the most uniform, so the percent change in MWF is perhaps low, which may not be clear based on just the mean difference maps

Thoughts This is more data than someone can humanly process, need to identify key regions Unsupervised exploratory data mining techniques could be worth pursuing, since our outcomes of EDSS and ΔEDSS are problematic – Goal here is to find patterns in the data rather than trying to predict an outcome