Principles of ROI volumetry

Question 1

What is Region of interest (ROI) volumetry?

Answer 1

1. Measuring the whole volume of a structure/structures

2. Can measure the whole brain but more usually select ROI (hippocampus, caudate nucleus etc)

Question 2

What are the ROIs informed from other studies?

Answer 2

1. Psychological deficits
2. Pathological features
3. Other imaging studies (PET, diffusion, perfusion)
4. Radiological features

Question 3

What are examples of ROIs?

Answer 3

1. Whole-brain volume
2. Lateral ventricular volume
3. Caudate volume
4. Putamen volume

Question 4

What may explain differences in subject group volumes?

Answer 4

1. Age
2. Gender
3. Across groups same scanner or proportion of scanners
4. Normal variation
5. Same rater or same proportion of raters was used across groups
6. Head size

Question 5

What is Longitudinal ROI studies?

Answer 5

INDIRECT delineate first and second structure and calculate change as vol1-vol2
DIRECT perform image subtraction method

Question 6

What is longitudinal studies used for?

Answer 6

Pathology or clinical neuroimaging

Question 7

What does cross-sectional ROI studies look at?

Answer 7

Hippocampal volumes in control and Alzheimer’s

Question 8

What are potential source of bias for longitudinal studies?

Answer 8

1. Are the intervals similar across disease groups
2. Were raters consistent across disease groups and across time
3. Were scan acquisitions consistent across disease groups and across time

Question 9

What is the cingulate cortex?

Answer 9

A region of increasing interest in Hungtington’s disease

Question 10

What is Anatior cingulate cortex (ACC) involved in?

Answer 10

Emotional processing and performance evaluation and optimisation

Question 11

What is posterior cingulate cortex (PCC) involved in?

Answer 11

Visuospatial orientation and spatial working memory

These are impaired in premanifest HD and early HD

Question 12

What is observed in preHD?

Answer 12

Both increased and decreased cingulate activation as well as reduced connecitivty

Question 13

What is TRACK-HD study?

Answer 13

A well characterised cohort of early HD and gene-positive premanifest subjects as well as healthy controls all of whom have undergone a large battery of assessment including 3T MRI

In addition to assessing disease-related volumetric differences, we examined the clinical correlates of these abnormalities

Question 14

What was the volumetric measurement of TRACK-HD study?

Answer 14

Cingulate segmentation followed a previously validated protocol,15 in which volume was defined as the gray matter of the cingulate gyrus including the retrosplenial cortex. The posterior limit of the cingulate gyrus was taken as the splenial sulcus, the superior and anterior limit as the bottom of the cingulate sulcus, and the inferior limit as the corpus callosum. Segmentation was performed in the sagittal view using a semiautomated method: consistent thresholds of 51% and 104% of the mean brain intensity were applied to generate an initial outline, followed by manual editing where necessary

Question 15

What is the Circle-Tracing Task?

Answer 15

The circle‐tracing task21 is primarily a task of visuomotor integration, performance monitoring, and evaluation. Subjects were required to trace a circle presented on a horizontally positioned monitor as quickly and accurately as possible. Initially, the subject was able to watch his or her performance directly, whereas in the indirect condition, the subject’s hand was obscured by a screen, and the performance was projected to a second, vertically positioned monitor. The error rate was determined as the number of deviations outside the outer limit of the target region of the circle per rotation for the indirect condition

Question 16

What was the statistical analysis for the TRACK-HD study?

Answer 16

Differences in head size between subjects were accounted for by standardizing cingulate volumes to the arithmetic mean TIV, using the cACC and PCC segments. The rostral segment was not used because the paracingulate added considerable variability to this volume. Specifically, the logarithms of cACC plus PCC volumes in the control group were regressed against the logarithms of the TIVs to establish the normal relationship between TIV and cingulate volume.

Question 17

What was the group difference in cingulate morphometry?

Answer 17

On average, the cingulate cortex was approximately 10% smaller in early HD subjects than in control with mean cingulate volume in the preHD subjects lying between these 2 groups

Question 18

What was the conclusion of this study??

Answer 18

the results of this cross‐sectional study suggest that the cingulate undergoes disease‐related structural degeneration that may initiate in premanifest HD but is clearly evident by early‐manifest disease. However, the cingulate is a complex and heterogeneous structure that is relatively insensitive to premanifest pathology when compared with structures such as the caudate. Consequently, we conclude that the cingulate would not provide a suitable biomaker of HD progression.

Question 19

What is the three main operational steps of the segmentation process are summarised here and presented in detail in the following section

Answer 19

Initialisation:
•
Definition of a bounding box delimiting the region of interest (ROI) including Hc and Am;

•
Placement of one seed in each structure (Hc and Am).

Alternate deformations:
•
Resetting of landmarks to NULL set.

Question 20

What might affect our ability to accurately measure a structure from an image?

Answer 20

Scan related: acquisition, artefacts: motion, inhomogeneity.
Subject related: hydration status of subjects
Rater related: intra and inter rater variability
Difficulty is often no gold standard

Question 21

estimating accuracy and reproducibility

Answer 21

Anatomical accuracy of segmentation… i.e. has my perfect hippocampal volume completely overlaid what is thought to be hippocampus?

Does my volume reflect what I drew previously or what someone else drew?

Important to not just consider overall volume, but actual overlap in regions – for example, two methods might give similar numerical volumes, but may not accurately overlap the actual region

Question 22

How to segment an ROI manually?

Answer 22

Can be very precise and accurate
Expert raters with high levels of repeatability
Sensitive
Often not reproducible by other research groups
Not feasible for small research groups or large studies
Time consuming (training and segmentation)

Question 23

How to segment an ROI via automated tools?

Answer 23

Reproducible 
Fast 
Often very reliable
Practicle for large studies
Do not require expert knowledge
Can be inaccurate 
Often particularly poor on patients with pathology – either under- or over-estimating

Question 24

What is automated segmentation of ROIs?

Answer 24

Voxel-based morphometry (grey and white matter segmentation)
FSL FIRST (cortical and subcortical volumes)
Freesurfer (cortical thickness measures and ROI volumes)
BUT can be a ‘black box’
Can be insensitive to longitudinal change
May introduce bias or artefacts (see Hobbs et al, Mov 2011, 26(9):1684-90)

Question 25

What is automated segmentation of ROIs?

Answer 25

Automated tools can be essential for small research groups or for large datasets
They do not require as much expert knowledge and training
However, they can introduce systematic errors – especially in participants with neuropathology
Testing of a software with quantitative and qualitative evaluation is essential – these tools perform differently on different software
Can give very good results and save a lot of time/effort when this is done
But they can introduce false negatives/positives if this is not done..!

Question 26

What are some implications for applying an inaccurate or unreliable segmentation tool (either automated or manual) within a clinical trial whereby caudate volume is a primary outcome?

Answer 26

If the drug slows or speeds up caudate change this might not be detected
Equally, the software could suggest changes where there is none
This could either under or over emphasise the effect of a drug
Resulting in either failed clinical trial OR an approved drug that is not actually changing the brain
Direct impact on the lives of patients
Caudate volume is used in Huntington’s disease clinical trials – this is a real risk!