Optimization and validation of small bowel water content estimation using MRI

1
Optimization and validation of small bowel water
content estimation using MRI

• C.L. Hoad1, L. Marciani2, S.Foley2, J.J. Totman3,
  J. Wright4, D. Bush4, E.F. Cox1,E. Campbell2,
  R.C. Spiller2 and P.A. Gowland1

1 Sir Peter Mansfield Magnetic Resonance Centre,
School of Physics and Astronomy, University of
Nottingham, UK, 2 Wolfson Digestive Diseases
Centre, Nottingham University Hospital,
University of Nottingham, UK, 3 Brain and Body
Centre, University of Nottingham, UK, 4
Department of Surgery, Nottingham University
Hospital, University of Nottingham, UK
2
Background

• Small bowel water secretion and absorption play a
  critical role in GI processing of food
• Water flow can be up to 9 l/day 1,2
• Current techniques are invasive and limited
• The ability to measure the water content of the
  small bowel both serially and non-invasively
  would enable new insights in GI physiology and
  pathology (e.g. Celiac, IBS) and action of drugs
• MRI provides a unique tool to investigate this
  area

1 Spiller R.C. Gut 35S5-S9, 1994. 2 Dunlop
S.P. et al. Am J Gastroenterology 1011944-1944,
2006.
3
Background

• Previous work imaged changes in small bowel water
  content upon addition of bran to a test meal
  ISMRM 2006 p840

4
Aim

• How accurate are these small bowel water content
  measurements?

5
Aim

• How accurate are these small bowel water content
  measurements?

• To optimise and then validate the small bowel
  water content measurements using naso-duodenal
  bolus infusions in healthy volunteers.

6
Methods

• Volunteers
• 18 healthy volunteers (9 M, 9 F), fasted
  overnight
• Intubation
• Gel lubricated 14fg (4.7mm diameter) Salem sump
  tube
• Tube position verified in duodenum using MRI

7
Methods

• Volunteers
• 18 healthy volunteers (9 M, 9 F), fasted
  overnight
• Intubation
• Gel lubricated 14fg (4.7mm diameter) Salem sump
  tube
• Tube position verified in duodenum using MRI

MIP of MRCP data segmented to show intubation
tube and surrounding anatomy
8
Methods

• Scanning
• Philips 1.5 T Achieva Scanner
• Coronal TSE sequence (MRCP)
• TEeff 320ms, TR8000ms, 24 slices, thickness
  7mm, FOV400mm, recon matrix 512x512
• Baseline scan, followed by 8x40ml boluses of test
  solution
• 145mmol/L NaCl and 15mmol/L Mannitol in water 1
• MRCP sequence acquired immediately after each
  bolus
• Total infusion experiment lt 15 mins

1 Spiller RC et al. Clin Sci (Lond) 67 117-20,
1984
9
Methods

• Analysis

Optimisation
Validation
Data not used
First 6 Subjects
Following 9 Subjects
3 Subjects 1 Subject withdrew 2 Subjects data
rejected 1 intubation tube failed to reach
duodenum test solution only in stomach 1
experiment took too long absorption/propulsion
10
Methods

• Analysis - Optimisation

11
Methods

• Analysis - Optimisation

Generate CSF level Scsf from histogram of CSF
12
Methods

• Analysis - Optimisation

Generate CSF level Scsf from histogram of CSF
Segment small bowel data using a threshold (Sth)
of 0, and save data for further analysis
13
Methods

• Analysis - Optimisation

Generate CSF level Scsf from histogram of CSF
Segment small bowel data using a threshold (Sth)
of 0, and save data for further analysis
Generate new Sth from different multiples of Scsf
(e.g. 1.5, 2, 2.5 Scsf etc) and calculate small
bowel water content volumes
14
Methods

• Analysis - Optimisation

Generate CSF level Scsf from histogram of CSF
Segment small bowel data using a threshold (Sth)
of 0, and save data for further analysis
Generate new Sth from different multiples of Scsf
(e.g. 1.5, 2, 2.5 Scsf etc) and calculate small
bowel water content volumes
Plot graphs of measured MRI volumes vs infused
boluses for different multiplying factors of Scsf
15
Methods

• Analysis - Optimisation

Generate CSF level Scsf from histogram of CSF
Segment small bowel data using a threshold (Sth)
of 0, and save data for further analysis
Generate new Sth from different multiples of Scsf
(e.g. 1.5, 2, 2.5 Scsf etc) and calculate small
bowel water content volumes
y1.0083x-4.2948 R20.995
Plot graphs of measured MRI volumes vs infused
boluses for different multiplying factors of Scsf
Find multiplying factor which gives slope of
graph closest to identity.
16
Methods

• Analysis - Optimisation

Develop algorithm to do this automatically
Generate CSF level Scsf from histogram of CSF
Segment small bowel data using a threshold (Sth)
of 0, and save data for further analysis
Generate new Sth from different multiples of Scsf
(e.g. 1.5, 2, 2.5 Scsf etc) and calculate small
bowel water content volumes
Validation study
Plot graphs of measured MRI volumes vs infused
boluses for different multiplying factors of Scsf
Find multiplying factor which gives slope of
graph closest to identity.
Retain in software
17
Results

• MIPs of segmented data at threshold Sth for a
  single volunteer

18
Results

• Validation data from 9 volunteers (N72
  measurements)
• Bland-Altman plot calculated 95 limits of
  agreement for the measured data was from 16
  below to 22 above infused volume
• Standard Deviation of differences was 10

19
Results

• Inter-observer variability of SBWC measurement
  (Bland-Altman 95 limits of agreement)
• 3 observers (N20 test data sets)
• -15 to 17 (maximum limits between all observers)
• Intra-observer variability
• 1 observer, 2 separate analyses (N20 test data
  sets)
• -4 to 3 (95 limits of agreement, Bland-Altman)

20
Conclusions

• Technique is reliable method for assessing small
  bowel water content non-invasively with maximum
  differences of 22 of volume measured. Standard
  deviation of differences 10.
• Errors from
• manual segmentation
• partial volume of fluid
• setting threshold level Sth
• Intra and inter-observer variability has a
  smaller error than actual volume

This work was funded by the Institute of
Neuroscience, University of Nottingham
21
Applications

• Changes in small bowel water content
• ingestion of foods / calorie content ISMRM 2007
  Talk 893
• disease (Celiac, IBS) ISMRM 2007 Talk 894
• mode of action of drugs (5-HT3 antagonists)

22
Applications

• Ondansetron Study

Double-blind, placebo controlled, randomised,
2-way cross-over study aimed to assess the
effects of Ondansetron in 16 healthy volunteers
Fasting small bowel water content
62
Two-tailed Students plt0.0001