Automated Diagnosis of Gait Abnormalities

1
Automated Diagnosis of Gait Abnormalities

• Chris Kirtley MD PhD
• Associate Professor
• Dept. of Biomedical Engineering Catholic
  University of America Washington DC, USA

2
Clinical Gait Analysis web-site
www.univie.ac.at/cga
3
Case of the Week
4
CGA Case of the Week

• Online discussion of cases presented on CGA page
• Highlights differences in interpretation and
  clinical opinion

5
Need for Autodiagnosis

• Complex and interdisciplinary nature of gait
  analysis
• Need for standardization in data interpretation
• Education and Training
• Collaborative research

6
Motion Toolbox
Kirtley C Smith RA (2001) Application of
Multimedia to the study of Human Movement.
Multimedia Tools and Applications 14 (3)
259-268.
7
Basic Principle

• Compare gait curves of patient against database
  of normative data
• Mean 1 standard deviation normally used to
  define normal range

8
Risks
9
Bayes Theorem
10,000 subjects

• Incidence of false positives and negatives
  depends on sensitivity, selectivity and (most
  importantly) the prior probability of an
  abnormality
• Emphasizes the importance of clinical history and
  exam in diagnostic process

Have disease (1) Normal 10,000/100
100 9,900
True False True False ve
-ve -ve ve 0.95100
0.959,900 95 5 9,405
495
Accuracy 95/495 only 20!
10
Definition of Normal Range
11
Need to Tighten Normal Range

• Match normative data as closely as possible to
  patient subgroup (age, gender)
• Calculate appropriate normative data for speed
  (or stride length, cadence) of patient

Kirtley, Whittle Jefferson (1985) Influence of
Speed on Gait Parameters. J Biomed. Engg. 7
282-289
12
Dimensionless Normalisation
Gait variables normalised to dimensionless units
to compensate for the effect of size and
mass Hof, 1996 Stansfield, 2001, 2003
13
Maturation of Speed
14
Maturation of Stride Length
15
Maturation of Cadence
16
Temporal Parameters

• Stance Duration -7.953s 69.602 ?
• -7.685s 69.519 ?
• Double Support -7.778s 19.636 ?
• -8.335s 20.168 ?
• (s walking speed in statures/s)

Rosenrot P, Wall JC, Charteris (1980) The
relationship between velocity, stride time,
support time and swing tome during normal
walking. J Hum Mov. Stud. 6 323-335.
17
Effect of Normalization
After normalization, effect of speed is shown to
be consistent Stansfield BW, Hillman SJ,
Hazlewood ME, Lawson AM, Mann AM. Loudon IR, Robb
JE (2001) Sagittal Joint Kinematics, Moments, and
Powers Are Predominantly Characterized by Speed
of Progression, Not Age J Paed Orth 21403-411.
18
(No Transcript)
19
Effect of Speed
20
Linear Regression

• Linear regression of each point across gait cycle
  for each variable at range of walking speeds
• Normative curves can then be plotted appropriate
  to the patients speed

21
Joint Kinematics

• Almost all regressions had good to excellent
  correlations
• Failed regressions always occurred when variable
  was approximately zero or constant

22
Joint Moments
23
Joint Powers
24
Electromyography
Hof (2003) Data on CGA site ltwww.univie.ac.at/cga/
datagt
25
The Rules
26
Spot Value Knee at Initial Contact
27
AverageAnkle Angle during Swing Phase
28
Positive IntegralAnkle Push-off Power
29
(No Transcript)
30
Induced Acceleration Analysis

• Means by which to determine the effect
  (acceleration) caused by a given muscle
  contraction
• Potential to remove subjectivity from gait data
  interpretation

Kepple, T. Siegel , K., Stanhope, S. (1997) The
use of two foot-floor models to examine the role
of the ankle plantar flexors in the forward
acceleration of normal gait. Gait and Posture, 5,
172-173.
31
Subject 1 Knee ControlWeakness in knee
extensors (3/5) and ankle plantarflexors (unable
to stand on tiptoe)
Knee extensor sources
150
100
50
rads/s/s
0
-50
Knee flexor sources
-100
Total

Ankle

Knee

Hip
Gravity
32
Primary source of knee extension
hip extensor moment
33
Subject 2 Knee Control Weakness in left (2/5)
and right lower limb (4/5)
Knee extensor sources
100
50
rads/s/s
0
Knee flexor sources
-50
-100
Total
Ankle
Knee
Hip
Gravity



34
Primary source of knee extension
contralateral ankle plantarflexor moment
35
Further Work

• Develop rule files for various conditions
  (amputees, stroke, Parkinsons etc.)
• Refine rules by sharing and multi-centre
  validation
• Higher level diagnosis

36
Please get involved!
www.univie.ac.at/cga/diagnosis