Next: Belief Network Theory
Up: No Title
Previous: Introduction
Subsections
Human Ambulation and Falls
The need for a system capable of detecting falls - especially falls in the
elderly - can be justified on a number of points. Firstly, if such a system
were to be made available to the public, it should be able to benefit the
patient by providing warnings or alerts to care-givers or medical staff if the
patient were to suffer a fall. In the case of elderly patients, who may not be
able to recover from a fall, this could help prevent further injury or medical
complications arising from lack of attention. Secondly, the physiologist or
medical staff interested in the well-being of the patient would most likely be
interested in a record of the number and frequency of falls their patient has
suffered, as an indication to their general stability and requirement for
further or a different type of assistance. A patient may not report falls to
their doctor out of embarrassment or simple forgetfulness (known as the
reporting problem). Such a system could provide not only an independent
record of falls - but also near-falls or stumbles, which may then be used to
determine the correlation between a falls and near-fall events for a patient
(see Section 2.2). Lastly, if such a system were to be
developed with remote monitoring/alerting capabilities and proven reliable, it
could provide a patient with a greater degree of personal freedom,
independence and confidence, as the need for a dedicated care-giver would be
reduced to the times when the system detects (or predicts) a fall.
Before discussing the causes and mechanics of falls, it is necessary to first
understand normal human ambulation.
The methods and mechanics of human ambulation - the bipedal gait - is quite
unique in the nature, and is not found in the same form in other animals. The
human gait has been described as consisting of a cycle of `controlled falls',
which highlights the complexity of distinguishing between a fall or stumble
and normal, controlled walking [8].
Evolution of the foot, upright stance and bipedal gait are considered
important stages of human evolution. Among other things, human ambulation left
the hands free to use tools, encouraging development of an increased mental
capacity [9].
The details of a human's walking pattern can be as individual as a
fingerprint2.1. Some factors that may result in one
individual's gait being distinct from another include muscle strength, tendon
and bone length, bone density, visual acuity, co-ordination skills,
experience, body mass, centre of gravity, muscle or bone damage, physiological
conditions, and a personal walking `style'.
The finer details of gait may not only vary from individual to individual, but
could change for one individual at different times. Individual gait is
dependent on variables such as surface properties, confidence, mood, quantity
and quality of sensory information (e.g light levels, noise, distractions),
any chemical effects on the body (such as alcohol or pharmaceuticals), energy
levels, mental alertness and muscle fatigue.
Gait cycle
Although the human gait varies from individual to individual and for different
environmental states, some basic stages are common for all normal gaits. This
is represented in the gait cycle (Table 2.1), which
consists of a repeated sequence of events used by humans to achieve
ambulation. These events are presented with respect to one leg (the other leg
represented as `opposite') and can be grouped into specific periods of the
gait cycle, which in turn fit into two main phases; the stance and swing
(Table 2.1, derived from [8, page
27]). The stance period consists of the time a given
foot is on the ground and providing a means of support to the body. The
stance phases of both legs overlap, producing intervals in the gait cycle
where both legs are supporting the body, which is known as double
support. The swing phase covers the stage when one limb is off the ground
and swinging forward under the body, while the opposite foot is supporting the
body's weight (which is then in single support). The extremities of the
gait cycle are the `toe-off' and `foot-strike' events, also marking the
termination of the stance and swing phases respectively [8, pages
25-27] and [10, pages 73-76].
Table 2.1:
Normal walking gait cycle period and functions.
| |
Period |
% Gait cycle |
Function |
| 1 |
Initial double-limb support |
0-12 |
Loading, weight transfer |
| 2 |
Single-limb support |
12-50 |
Support of entire body
weight, centre of mass moving forward |
| 3 |
Second double-limb support |
50-62 |
Unloading (toe-off) and
preparing for swing (pre-swing) |
| 4 |
Initial swing |
62-75 |
Foot clearance |
| 5 |
Mid swing |
75-85 |
Limb swings towards front of
body |
| 6 |
Terminal swing |
85-100 |
Leg deceleration, preparation
for shifting weight to opposite limb |
|
The gait cycle is achieved by a precise, energy-efficient sequence of muscle
actions. The muscles and joints perform multiple tasks, including shock
absorption during foot-strike and deceleration, joint stabilisation (ankle,
knee, hip), loading response, and leg extension to raise the body's centre of
mass during the swing phase. The forward movement achieved during normal
walking makes efficient use of the body's forward momentum, which effectively
`falls forward' from the point where the supporting leg is fully extended and
the body's centre of mass is highest. The muscles of the opposite leg stop the
fall and absorb shock at the next foot strike, then the cycle repeats
[11].
The DBN models built during this project to monitor ambulation and recognise
falls rely heavily on recognising stages of the gait cycle. Characteristics of
a cycle, such as no support between single-limb support periods and small time
intervals between steps, are used to recognise stumbles, falls and different
ambulation modes such as running or jogging (Chapter 8).
Fall and ambulation monitoring
A fall, otherwise known as a fall event, has been defined as
``An event which results in a person coming to rest inadvertently on the
ground or other lower level, and other than a consequence of the following:
sustaining a violent blow, loss of consciousness, sudden onset of paralysis,
as in a stroke, or an epileptic seizure'' [12]. For the purpose of
this project a `fall' will entail someone loosing balance or tripping, and
dropping to the ground during ambulation. As this project focuses mainly on
stepping patterns and gait, the definition used here does not include falling
from heights (such as from ladders) or other events that that don't directly
involve normal upright walking or running. A stumble is also used to
describe the event when a subject momentarily loses balance, but recovers and
does not actually fall. A stumble is usually characterised by a quick
correctional step (see Section 2.3.2).
A fall event may occur due to a combination of factors. Some may be related to
the operation or malfunction of bodily systems responsible for detecting and
maintaining balance (e.g neurological, sensory and motor systems). Other
factors may be environmental and outside the control of a fall victim, such as
slippery surfaces, obstructions, sudden impact or unexpected movement of the
surface they are standing on. In any case, a normally functioning body
generally tries to avoid injury to itself by automatically applying tactics to
avoid or at least minimise damage. In some cases such as loss of consciousness
or loss of motor control, the body's system may not be able to provide such
protection.
The main task of balance (while standing or ambulating) is to maintain the
body in an upright position by aligning its centre of gravity (COG)
over its base of support (BOS). The COG, also known as centre of mass,
is the point of a rotating body that would move in the same path as a single
particle subjected to the same forces [13]. In the case of most
humans, the body's COG usually lies at the centre of the pelvis.
The base of support is the minimum area enclosing the body's contact with the
ground. Therefore, while standing the BOS is the area enclosing the soles of
the feet (or shoes). A smaller BOS gives a smaller area for the COG to be
aligned over, and hence the body can be considered less stable. Similarly,
when sitting or lying down, the BOS is much larger, so the body is much more
stable.
If the intersection point of the support surface and the vector starting from
the COG and pointing in the direction of gravity moves outside the BOS, the
body detects a reduction in balance and attempts to compensate (see Section
2.3.2).
Balance systems
The human body incorporates some ingenious methods to maintain balance. This
is no simple task, given we tend to prefer bipedal support, as opposed to the
more stable quadrupedal popular among many other animals. Failure or confusion
in any of our balance systems may lead to stumbles or falls, which can then be
detected. Although this project cannot directly assess the quality of an
individual's balance mechanisms, particular problems may be inferred from its
detection of fall events and other ambulation information.
Three bodily sensory systems are primarily responsible for detecting and
maintaining balance. These are the visual system, the vastibular system and
proprioceptive system. These systems work together to determine the body's
location, surroundings, orientation, current movement, limb position, and
overall stability. The central nervous system monitors the sensors and
initiates required involuntary adjustments, or reflexes, via the motor control
systems [10, page 61].
Vision is one of the more complex and heavily relied upon senses which also
plays an important role in balance and fall prevention. Vision is used to
detect obstacles, predict slippery or dangerous surfaces where extra caution
should be exercised, detect changes in walking surface level or slope, and
provide a means to plan a safe route of travel. The visual system can also be
used to determine distance using depth perception, speed, position, height,
orientation of both the viewer and any objects within their field of view
[10].
Vision relies on certain amounts of light to operate - too little and objects
cannot be seen properly, increasing risk of tripping and falling. Too much
light (e.g. glare) can have similar effects, as objects are obscured by bright
light sources or reflections. Obstructions representing potential problems can
also be obscured by shadows and particular combinations of colours or textures
that either reduce depth perception or confuse the visual system.
Vision also relies on higher neurological reasoning to process and recognise
images, and determine which objects or surfaces in the environment could
hinder or aid balance. This information can then be used to decide what
actions to use when walking on or around that area. For example, when we see
icy path, we tend to slow down, pay more attention to balance and `tread
carefully'. Vision also provides the brain with visual references like the
horizon, ground plane or landmarks, which can be used to determine
orientation or heading.
An individual's reliance on their visual system for balance can be tested by
asking a patient to walk and stand with then without their eyes closed (the
Romberg test). A patient experiencing difficulty maintaining balance
without the use of their vision could suggest an over-reliance on vision
resulting from problems in the other balance systems [14, page
85]. Similar diagnostics could be performed using a blindfolded
patient connected to an ambulation monitoring system (such as the one
implemented here), which may be able to provide more detailed information
relating to the patient's status.
The vastibular system is responsible for detecting orientation and movement
of the head, and as such provides the basic sense of `balance'. Its two
primary sensory organs are the utricle to sense head orientation (with
respect to gravity) and the semicircular canals to detect rotational
movement. Both of these organs are situated adjacent to the cochlea, and form
part of the inner ear [15,16, page 117]. Effects such
as `dizziness' can be attributed to residual or abnormal motion of the fluid
in the semicircular canals following rotational motion. The vastibular system
can also be affected by chemicals such as alcohol which may change the
properties of the fluids used in the utricle and semicircular canals,
adversely affecting balance.
As well as providing a sense of balance, the vastibular system stabilises the
eyes during movement of the body and/or head. When walking, the entire body is
subjected to vibration and movement, but the vastibular system allows the
field-of-view to remain fixed and focused on objects or the path ahead. This
is accomplished by adjusting eye orientation in an equal and opposite
direction to head movement [10]. Failure of the vastibular
system would result in a constantly changing field of view, hindering vision
and therefore balance.
The proprioceptive system (proprioceptive feedback) consists of receptors in
muscles, tendons, joints and the feet used by the brain to determine the
orientation and position of limbs. In the case of lower leg and feet
receptors, the orientation of ankles and pressure distribution on the feet can
be used to determine the orientation of the surface the body is standing on,
as well as detect vibrations via the feet. Sensing the position of joints is
used to reduce ``postural sway'' and keep the body stable [10].
Motor and musculoskeletal balance response
When a balance system detects a potential problem, the motor and
musculoskeletal systems are employed to adjust posture, keep the COG point
within the BOS and remain upright. These motor responses may involve small
adjustments such as postural sway (ankle strategy), adjustment to the knees
and hips (hip strategy) and a corrective step (step or stumbling strategy). In
general, a larger balance correction results in a more pronounced reaction
including a greater number of body parts situated further from the feet
[10].
is used to correct small discrepancies in
balance. The feet remain planted and the body moves about the ankle joints
like a pendulum to realign the COG and BOS. Ankle strategy corrections
generally occur in flexors and muscles starting at the feet and moving upwards
towards the trunk.
performs the same realignment tasks as ankle
strategy, but on a larger scale. Hip strategy is commonly used where the BOS
is reduced, for example when one foot is in front of the other, as if walking
on a beam. As opposed to ankle strategy, the order of muscle reactions starts
at the trunk and moves down to the knee and ankles.
is used for large corrections where the COG's
downwards vector ground intersection point moves outside the BOS. This usually
involves a corrective step, stumbling or hopping. This response is actually
very similar to normal controlled walking, as the when moving forward, the COG
moves outside the BOS and a forward step is taken. This activity can be easily
detected through sudden drops in foot-strike to foot-strike timing intervals
and abnormalities in the gait cycle (Section 8.6).
Additional corrective reactions include the rightening reflex, where
upper limb, head and trunk movements are employed to sustain balance. This
could also conceivably be used to detect falls by attaching sensors to arms or
hands, although such methods may be overly intrusive and may not detect
unstable situations where no rightening reflex is performed (such as passing
out).
All the above reflexes tend to be corrective, as they compensate for
instability after it has occurred. Higher level techniques involving some form
of forward planning are used to perform preventive adjustments, usually
drawing on experience, vision and/or reasoning. Examples include slowing down
and `watching your step' on uneven or slippery surfaces, using extra support
such as hand-rails, and attempts to increase base of support area (e.g.
sitting) or adjusting the COG (e.g. ducking or outstretched arms).
The primary motivation for investigating fall detection for the benefit of the
elderly - as opposed to other groups - is that the elderly are generally
considered an `at-risk group' [17, page 10], more likely to suffer a
fall [1,18], and if they do fall, are more likely to do
themselves serious damage due to increased frailty [19,17]. A
system to detect and alert others to the occurrence of a fall in an elderly
person cannot realistically be expected to prevent falls before they happen,
but could at least help reduce the effects of any injury through prompt
medical attention and establish whether an individual has a propensity to
falling which may require more active assistance.
In Australia, the prevention of falls in older people was identified not only
as one of the causes of injury requiring the most attention, but also one of
the categories with smallest amount of available information regarding
appropriate intervention [17, pages 34-35].
Studies by Day et. al. [18] on a population of 5101 elderly cases
presented to metropolitan hospital emergency departments have shown that the
admission rates of the elderly was much higher than younger adults and more
females suffered from injuries than males2.2. Falls were the most common cause of injury (66%) and
injury most commonly occurred in the home (96%). In the home, falls
accounted for 69% of injuries and 87% of injuries in residential
institutions.
According to the Australian Institute of Health and Welfare [17],
falls were the third largest cause of of reported injuries
(15%)2.3, and the largest cause of hospitalisation (33%) in 1997. During
1995 there were 827 fall-related deaths and 57,934 fall-related
hospitalisations, with the elderly representing the largest proportion of
these numbers [20, page 5].
Statistics show that fall-related fatalities increase with age; during 1996 13
per 100,000 in the 70-74 age group suffered fatal falls, but this number
increases to over 200 per 100,000 in the 85 and over age group [20].
According to a report by the National Injury Prevention Advisory Council
[20], of the more than $3,000 million spent on falls in 1995,
$1,083 million was attributable to falls in the elderly. With the expanding
elderly
population, these costs are likely to increase.
Reasons for the higher proportion of falls in the elderly are usually at least
partly the result of natural time-degradation of facilities used in balance
(Section 2.3.1), motor control, cognitive skills and
ability to handle (or detect) environmental changes such as lighting and
surface quality [1,21]. The higher probability of
injury in elderly people who have suffered a fall can be attributed to
factors such as low bone mass density, smoking, weight, and disease
[21].
Besides age, other factors which appear to effect the probability of a fall
occurring in an individual include gender, living arrangements, level of
fitness and general health.
Below 70 years of age, some statistics show that gender doesn't appear to be a
large contributing factor to the occurrence of falls. For ages over 70
however, the female population appears to be at a higher risk of suffering
more falls than males [19,18].
The largest proportion of the falls tend to occur in private and nursing
homes, followed by outside the home, during activities such as gardening
[19]. Fall hazards within the home include steps and stairs, chairs,
floor surfaces, beds and ladders. Outside the house, uneven surfaces such as
foot-paths and steps can contribute to falls [18].
Some forms of medication have also been identified as increasing the risk of
falls [17, page 61]; this may result from confusion on the patients
part resulting in inappropriate use and side-effects.
The physiological and medical results of a fall range from grazes, cuts,
sprains, bruising, fractures, broken bones [19], torn ligaments,
permanent injuries and death. The most common fall-related injuries are
lacerations and fractures, the most common area of fracture being the hip and
wrist [18]. Damage to areas such as the hip and legs can result in
permanent damage and increased likelihood or further falls and injuries.
Falls in the elderly can produce social and mental as well as physiological
repercussions. Anguish and reduced confidence in mobility can complicate
problems and even increase the probability of falls and severity of
fall-related injuries from lack of exercise and tensing of muscles. Dr. Ian
Brown hypothesises that those who have suffered a fall and resulting reduced
confidence in their mobility may tend to `override' some aspects of their
balance systems and its reactions, placing heavier reliance on senses such as
vision and touch, which may not be as reliable as the individual suspects [22].
The following related works focus on monitoring and detecting a wide range of
problems related to elderly living, falls being one of the major factors. It
should be noted that unlike this project, those described below attempt no
higher-level recognition of the data they are collecting, instead relying on
reasonably simple event triggers and manual human post-evaluation of data.
The CSIRO is researching and developing a `home telecare' system that aims to
remotely detect all sorts of problems linked with elderly living, to help
increase independence for the person involved, while maintaining medical
monitoring and support at the same time. The ``hospital without
walls'' [23] scheme uses radio-telemetry technology and a wide
variety of sensors for collection of data including acceleration (for falls),
cardiovascular status, noise, blood pressure and body position.
The system is confined to a single unit similar to the Compumedics Seista, but
built specifically for the purpose, producing benefits such as longer battery
life. Heart rate is collected by detecting ``R'' waves collected from ECG
recordings, but measuring blood pressure is more complicated, requiring the
standard and usually bulky inflatable blood pressure monitor. During a meeting
with Laurie Wilson from this project some other data collection methods were
mentioned, including sensors embedded on clothing, and the application of
devices used in sport to measure heart rate and other body-status
information.
Fall detection is performed using accelerometers originally designed for
deploying airbags in cars. The two accelerometers are contained within the
hip-mounted device, and can measure acceleration in three orthogonal
directions. According to Wilson, taking acceleration measurements near the hip
area can provide useful information - even if single footsteps are not
detected - as the hip is near the body's centre of mass. Detection of
occurrence and direction of falls can be performed via appropriate signal
analysis of the accelerometers.
Doughty [24] discuss the use of `smart' sensors and simple logic
rules to detect events such as falls and artifacts of dementia including
wandering and rummaging. Like the CSIRO Hospital Without Walls project, this
system uses a range of sensors on the body, but also incorporates various
sensors distributed around the living area. These cover detection of
environmental as well as personal status - such as leaving the gas on,
rummaging through drawers, wandering, running water and excessive heat. Body
sensors include a fall detection system dubbed `FRED' (Fall Response Emergency
Detector) which can be worn on the waist or in pocket. All these sensors are
combined home network and external communications to transmit alarms and
general status information.
As a result of investigations [20,17], the Australian Government
implemented the 1997 Aged Care Act [25] and increased funding
[26] to the area of aged care, resulting in a number of
government-based projects and initiatives.
Next: Belief Network Theory
Up: No Title
Previous: Introduction
Daniel J Willis
2000-10-23