Short course - ACTA (2023)
Weekly activities
Monday: Discussion with Dr. Lin (~ 1.5-2 hours). Please check this website for our preliminary schedule of the topics of discussion.
Tuesday: You are welcome to join our MSc/PhD seminar (English discussion, 2 hours) and my course (Research Design and Meta-analysis (1.5-2 hours).
Wednesday/Thursday: MRI study and oral assessment (at the MRI lab) (1.5 - 2 hours per participant). The study will be conducted by me and our RA.
Friday: Data analysis
Special notice:
You do not need to have your own laptop or PC for the study because I will have one ready for you - this is because of security reasons (lab data should be processed using our own laptops).
You do not need to stay in the lab all the time! Because the seats are very limited in the department building. I would suggest you spent more time reading and doing analyses in the university library, where is a better place for study and closer to your dormitory :)
However, please let me know if you have a new schedule deviated from the weekly schedule. For example, if you cannot participate in the MRI study on some Wednesday, please email me in advance.
Please create an account and gain access to Mendeley (https://www.mendeley.com) if you don't have one. We will need to use this cloud service for literature study.
9/04 - 9/08 Formulation of hypotheses: setting up research aims
Articles
Gonçalves, Schimmel, van der Bilt et al. (2021) Consensus on the terminologies and methodologies for masticatory assessment. J Oral Rehabil (https://pubmed.ncbi.nlm.nih.gov/33638156)
Zatorre, Fields, Johansen-Berg (2012) Plasticity in gray and white: neuroimaging changes in brain structure during learning. Nat Neurosci (https://pubmed.ncbi.nlm.nih.gov/22426254)
Instruction
The Gonçalves's paper summarizes some basic concepts about defining 'masicatory peformance'. Please keep in mind that there are many approaches to assess 'mastication' and the approaches may follow different definitions of mastication (e.g., mixing vs. cutting). You will find the early approach (Fig. 3) that Schimmel et al. used, and many different approaches derived from the concept of color mixing.
The Zatorre's paper is a very long one and you just need to focus on Fig. 1, which demonstrates the basic concept of study design of analyses of brain structure. You may need to focus on Fig.3, which tells you the biological significance of gray and white matter.
9/11 - 9/15 CAT12 image analysis + Gum images
Articles
Fischl & Dale (2000) Measuring the thickness of the human cerebral cortex from magnetic resonance images. PNAS (https://pubmed.ncbi.nlm.nih.gov/10984517)
Fjell et al. (2009) High consistency of regional cortical thinning in aging across multiple samples. Cereb Cortex (https://pubmed.ncbi.nlm.nih.gov/19150922)
Parisius et al. (2024) Towards an operational definition of oral frailty: A e-Delphi study. Archives of Gerontology and Geriatrics (https://www.sciencedirect.com/science/article/pii/S0167494323002595)
Practices
Run DICOM conversion (SPM)
Run segmentation (CAT12)
Visual check MRI segmented images (CAT12 / 'mri' folder)
Explore datasets of cortical thickness of ROIs (CAT12 / 'label' folder
Quantify the degree of mixing (VARG index) of gum images
Instruction
Try to find answers for the following questions from the articles by Fjell et al. (2009): (A) In young adults, what are the brain regions with greater / smaller cortical thickness? (B) In older adults, what are the brain regions with greater / smaller cortical thickness? (C) Are there any brain regions in which cortical thickness remains constant as age increases? (D) Are there any brain regions in which cortical thickness decreases as age increases?
Please read the Parisius's paper and focus on Table 2 (i.e., the components of oral frailty that reached a consensus among experts) - Can you figure out the brain's role in these components? (e.g., swallowing is about the reflex, and reflex may be associated with the central nervous system......)
9/18 - 9/22 CAT12 image analysis + Questionnaires
Articles
Zelig et al. (2019) The Eating Experience: Adaptive and Maladaptive Strategies of Older Adults with Tooth Loss. JDR Clin Trans Res (https://pubmed.ncbi.nlm.nih.gov/30931718)
Hsu et al. (2012) Evaluation of a self-assessed screening test for masticatory ability of Taiwanese older adults. Gerodontology (https://pubmed.ncbi.nlm.nih.gov/22295976)
Practices
Run Get TIV (CAT12)
Run Check Homogeneity (CAT12)
Translating the MAEQ
Quantify the degree of mixing of gum images using PhotoShop
Instruction
The whole article by Zelig et al. is inspiring because this may be the first article that investigated the experience of adaptation during eating. This article would be pivotal to your thesis Introduction and also Discussion.
For the Hsu's study, just focus on Table 2 (the chewing difficulty of different food types). The findings justified why we need to assess masticatory adaptation experience for 4 foods.
In terms of the 'Photoshop' analysis of gum images, please send me the jpg with its background removed (using the 'wand' tool?) - I will do the rest to see the effect of this manipulation.
9/25 - 9/29 CAT12 image analysis + Matlab
Articles (for a preview)
Ay et al. (2022) Comparison of FreeSurfer and CAT12 Software in Parcel-Based Cortical Thickness Calculations. Brain Topography (https://pubmed.ncbi.nlm.nih.gov/36208399)
Practices
Matlab script (extracting and concatenating data)
Presentation
Please have a short presentation about what you have learned from the articles listed between 9/4 and 9/22 - no more than 8 PPT slides ; )
In the slides, you can try answering the questions raised in 'Instruction', or bring your own questions that you felt confused when reading the articles.
Instruction
We do not have more additional work to do for this week. Please concatenate the data using the Matlab script (you can try doing this to other data, e.g. gray matter volume and white matter volume of ROIs in the same way.
The article by Ay et al. (2022) is for our next discussion. It is more 'technical' and you just need to have a quick browse, no need to read through it.
10/2 - 10/6 Clinical analysis
Articles (for a preview)
Joy et al. (2023) Alterations of gray and white matter volumes and cortical thickness in treated HIV-positive patients. Magn Reson Imaging (https://pubmed.ncbi.nlm.nih.gov/36265696)
MOCA: https://mocacognition.com/paper (here please look into the English version)
Practices
Bland-Altman plot (https://datatab.net/tutorial/bland-altman-plot)
Presentation (summary)
Instruction
We are heading to the second part of the externship research. More to do with the association between brain imaging (e.g., cortical thickness) and clinical (e.g., MP and questionnaire) data.
In addition to the variables of oral factors, I think we can put the scores of the cognitive test (MoCA) into analysis. Please fine the information of the latest version of MoCA.
(A) Neuroimaging basics: space orientation
(B) Neuroimaging basics: brain atlases
**Video learning: Tips for hypothesis testing of the clinical data (please find the video link in the email)
** Resource: Type I error [link] / Khan Academy
** Resource: p value [link] / Khan Academy
*Resource: The MNI brain and the Talairach atlas [link] / MRC Cognition and Brain Sciences Unit (2017) / University of Cambridge
*Resource: The Online Brain Atlas Reconciliation Tool (OBART) [link] / Brain Architecture Project
*Article: Tzourio-Mazoyer et al. (2002) Neuroimage [link]
9/25 - 9/29 Preprocessing
**Video learning: Imaging registration and transformation / (Introduction to Neurohacking In R) Johns Hopkins U via coursera [link] (15 min)
Tips:
Please ignore all the mathematics from that video... all you need is the lecture and the figures ; )
3:30 - This is a critical slide that shows the 'types' of registration.
Registration can be classified according to the methods of image transformation (or simply speaking 'the reshaping'). Rigid-body trans. consists of 6 degree of freedoms (or 6-ways) of reshaping: translation (x,y,z) and rotation (x,y,z).
Affine transformation = rigid body (6 d.f) + shearing (3 d.f) + scaling (3 d.f) therefore 12 d.f.
9:50 - All the above are linear transformation methods, by which an image is reshaped globally. And nonlinear transformation allows for local deformation.
So keep in mind that when doing registration, we start from linear trans. first and then nonlinear trans. next. Because we always need to match the global figures first, and then 'fine-tune' the local matching.
Also from this video, try to clarify the concepts between 'co-registration' and 'normalization'
Co-registration: when you match between two images from the same brain (e.g. to reshape my functional brain image to match with my structural brain image).
Normalization: to match the individual brains to a referential brain image, such as 'normalizing' subjects' structural brain to a template brain. If the template brain is in the MNI space, then after normalization, everyone's brain will be set in the same MNI space - and then we can compare between them.
**Video learning: The Multiple Comparisons Problem / Sprightly Pedagogue [link] (4 min)
Tips:
The value alpha (the 'alpha' in p<0.05=alpha) means the criteria for controlling type I error (or simply speaking 'false positive'). To be conservative, smaller alpha (or a more 'strict' criteria) would reduce the risk of false positive.
1:30 - Why cannot you do 'multiple rolling'? Note the difference between these two statements:
(A) I roll a dice for 1 time and there will be a '3' - the chance for hitting this hypothesis is 1/6.
(B) I roll a dice for 6 times and there will be at least a '3' - the chance will not be 1/6. It will be [1- (1- 1/6)^6] =0.665 !
Therefore, when you are doing multiple testings or multiple comparisons or simply 'multiple trials', just like 'multiple rolling', you actually increase the probability to hit the hypothesis.
Please note that: for each single trial, the hit chance is still 1/6. But for the whole set (10 times) of trials, the 'familywise' (that means 'the whole set') chance becomes larger.
And what about to roll it for 10000 times (and loudly claim that 'I got a '3' finally in the 9999th run) - that's very unfair.
3:30 - Therefore, we need to control for the familywise chance. That is why we need to control for the familywise chance for hypothesis testing. -- one of the methods is Bonferroni correction.
10/02 - 10/06 Statistical analyses
10/09 - 10/13 Hypothesis testing
** Resource: Partial Correlation Practice Problem [link] / Statistics (PSY 210 and ECON 261) at Nevada State College
To-dos
Group comparison: according to the subgroups (you decided based on posterior contact), compare the (a) masticatory performance and (b) age between the subgroups.
Correlation: assessing the partial correlation by controlling for the variable 'age'. So for each analysis, you will perform (a) Pearson's r, (b) Spearman's rho, and (c) the r controlled for 'age'.
Making x-y plots and bar charts (next time)
Appendix: Data analysis and How to make a chart
**Video learning: Tips to make a histogram, a x-y plot and a bar chart [link]
** Resource: Excel instruction for making a x-y plot [link]
** Resource: Excel instruction for making a bar chart [link]
** Resource: Excel instruction for the function 'countif' (for analyzing frequency and making a histogram) [link]
To-dos
For the key variables that you are analyzing (e.g. number of missing teeth, saliva flow rate, masticatory performance etc.), make a histogram for the individual variable to present the distribution of its scores.
For the analysis of correlation, present the results using a x-y scatter plot.
For the analysis of group comparison, present the results using a bar chart.
Tips: Regarding making a histogram...
Remember that you need to decide the range (i.e., the min and max values) first. The x-axis of the histogram will cover the full range of the score.
There is no rule how many divisions (or 'bins') to be presented for a histogram. If your scores are continuous (e.g., saliva flow rate), try to define an reasonable interval so that you can get around 6~8 divisions.
Remember that for all histograms, the y-axis represents 'frequency' (or 'counts'). Therefore, a histogram will help you to find out the the 'mode' of a distribution, which is the score mostly often appears in a set of data.