CBU Methods Research and Infrastructure Group

1
CBU Methods Research and Infrastructure Group

• Ian Nimmo-Smith

2
Personnel and Fields of Expertise

• Statistics and Mathematics
• Ian Nimmo-Smith Peter Watson
• Physics of Magnetic Resonance
• Christian Schwarzbauer
• Imaging Data Analysis
• Matthew Brett
• An irreducible minimum

3
The wider picture

• Expertise in other research groups
• Rhodri Cusack
• Ingrid Johnsrude
• Friedemann Pulvermüller, Olaf Hauk, Yury Shtyrov
• Commitment across the Unit to develop new and
  improved methods of research
• Supplemented by visitors, collaborations, grants

4
Strategy of the Methods Group

• Foster a critical understanding of existing
  methods
• Recognise and respond to the need to develop new
  or improved methods

5
Scope of the services offered

• Consultancy services
• Identify available methods and implement them
• Develop new methods
• Training
• Website
• Software
• Seminars

6
Statistical and Mathematical Modelling

• Meta-analysis of foci of activation
• Efficient estimation of pitch ordering of
  electrodes in auditory brainstem implants

7
Progress MR6.1 Meta-analysis of foci of
activation

• Three-dimensional spatial distributions compared
  across tasks and studies using a higher-dimension
  extension of the classical Kolmogorov-Smirnov
  two-sample test
• First meta-analysis of foci in terms of their
  co-ordinates

Applications to central executive function
(Duncan Owen) localisation of emotion (Murphy
Lawrence) hypofrontality in schizophrenia
(Hill al.)
Activation foci of Tasks using episodic memory
have a more frontal distribution than those of
tasks where cognitive demand is manipulated
8
Proposal MR4.4 Efficient ranking methods for use
with auditory brainstem implants
A good channel-to-electrode mapping is needed to
pass the correct speech information to an
appropriate electrode. With a Cochlear Implant,
the most apical electrode is in a low-frequency
(LF) place and the most basal is in a
high-frequency (HF) place. So there is a clear
channel-to-electrode mapping. With an Auditory
Brainstem Implant, the 21 electrode paddle is
placed on the Cochlear Nucleus. The
frequency-to-electrode map is unclear. So there
is no a priori correct channel-to-electrode
mapping.
Cochlear Implant
Auditory Brainstem Implants
High? Medium? Low?
High? Medium? Low?
High? Medium? Low?
High? Medium? Low?
9
Proposal MR4.4 Efficient ranking methods for use
with auditory brainstem implants
Combinatorial explosion! 3 electrodes have 3! 6
possible rankings. This can estimated with no
more than 3 paired comparisons (see tree) 21
electrodes have more than 5 x 1019 possible
rankings. This can be estimated with no more than
71 comparisons. (Steinhaus binary insertion
algorithm). Work in progress will reduce this to
66. (Ford Johnson merge insertion
algorithm). Significant improvement in time and
accuracy to assist neurosurgeon and
patient Future work will find effective ways of
incorporating prior information
Binary Insertion decision tree based on
sequential paired comparisons