Keywords
tongue–palate interactions - complete denture - speech production - swallowing - strain–gauge
sensors
Introduction
To study with good spatial and temporal accuracies the mechanical interactions between
tongue and palate in speech production and swallowing, we have developed a system
called PRESLA, which relies on the use of a palatal plate with embedded strain-gauge
sensors and innovates by using complete edentulous subjects[1] who have been using their complete denture for months and adapted to articulate
and swallow with it.[2] In the recent years, the PRESLA system beneficiated from numerous improvements,
including the design of specific acquisition chain, which enables high signal-to-noise
ratio, an increased number of sensors, specific data-processing programs, and an improvement
in the calibration device. This report briefly documents the technical aspects of
our experimental setup, and presents first results for fricative consonants and swallowing
experiments obtained from a single subject.
Materials and Methods
Participant
A 54-year-old completely edentulous male subject, native French speaker, served as
subject. He received all the required information about the experimental setup, the
protocol, and our research aims, and signed informed consent. His participation is
part of the pilot study associated with the ethical committee application 69HCL18_0103
promoted by the University Hospital of Lyon.
The subject had worn maxillary and mandibular complete denture satisfactorily for
3 years. An important issue in edentulous patients, which has been frequently addressed
in the literature, is the so-called “macroglossy,” a term abusively used to refer
to the apparent increase in the volume of the tongue in its anterior part. It actually
refers to the fact that in these patients the shape of the tongue is modified due
to the absence of teeth: the tongue spreads over the edentulous ridges, which induces
a reduction in the vertical dimension of the air channel and limits the freedom of
the tongue to move vertically. Our subject did not suffer from this problem. The clinical
evaluation of his oral cavity has revealed that by the time of the experiment he had
recovered nonpathological physiological parameters, in particular with a normal size
of the air channel between the tongue surface and the palatal vault, mainly due to
the correct restoration of the vertical dimension of occlusion and to an appropriate
thickness of the prosthesis, which preserves the freedom of the tongue to move vertically
and horizontally in the vocal tract. In addition, informal evaluations of his speech
production did not present any indications of speech disorders. In agreement with
these observations, this subject reported no history of medical conditions such as
speech impairments and dysphagia. He had also no history of hearing disorders. Hence,
we consider that our subject had fully adapted to his complete prosthesis and had
normal speech production and swallowing by the time of the experiment. Importantly,
we consider then that the data collected from this subject inform about normal speech
production and swallowing.
Subject-Specific PRESLA System
The original PRESLA system has been presented in our seminal paper published in Material Science and Engineering C.[1] In particular, details about the strain-gauge sensors and the instrumentation tools
can be found in Figure and sections 2.1.4 and 2.1.5 of this seminal paper (pages 836–837).
Additional information about the recent improvements in the setup is given in our
recent paper[3] published in Clinical Linguistics and Phonetics, in section “Material,” pages 57–61. We focus here on experimental aspects associated
with the investigation of speech production and swallowing in our subject.
Six strain-gauge transducers were inserted into the palatal surface of the duplicate
([Fig. 1B]). To determine the locations of sensors in the palatal plate, high-quality green
spray (Occlu Spray Plus, Hager & Werken GmbH & Co. KG, Germany) palatograms were done
on an accurate duplicate of the subject’s complete denture, for both speech and swallowing
tasks (one recording per swallowing task and per consonant; [Fig. 1A]).
Fig. 1 (A) Palatogram for fricative consonant /s/ on existing complete denture to determine
sensor position. (B) Sensors placed into the duplicate. The wires are placed inside the thickness of
the duplicate. Sensor surfaces flush with the palatal plate.
Small holes were drilled in the palatal plate to accommodate the sensors, without
altering the palatal vault of the complete denture. Each sensor was fixed on a plastic
disk to avoid the effect of strain in the resin, taking care to ensure the free movement
of the strain gauge. The sensing surface was aligned along with the palatal surface.
The wires connecting the sensors to the amplifier, placed within the thickness of
the base plate, exit from the vestibular side of the premolar area by the way of labial
commissure.
Calibration
Calibration of all the sensors was performed in a specially constructed setup we have
developed to have conditions similar to those of the experiment, especially as regards
the soft characteristics of tongue tissue, to measure tongue pressure. The calibration
setup is based on the concept of water column exerting a perfectly controlled pressure
on the sensor via a latex finger ([Fig. 2]). The experimental denture was firmly mounted on a fixed vice with a height and
an inclination enabling that the latex contacted the sensor at essentially right angle
and without exerting any stress on it. This configuration was considered to determine
the reference water height and reference sensor output associated with a zero-mechanical
pressure on the sensor. Mechanical pressure was then applied to the sensor by increasing
precisely the water height. The relation between the mechanical pressure and the output
voltage of the electronic measurement system (a Wheatstone bridge) has been shown
to be linear in the range [0, 5 kPa], and we assumed this linear relation to apply
over a larger range up to 50 kPa (see Mirchandani et al[3] for details).
Fig. 2 Global view of the calibration device (left); latex finger and vice holding the denture
(right bottom); and electric motor display showing the height of the water (right
top).
The accuracy of the pressure measurement provided by PRESLA was evaluated with the
water column across five repetitions of the measure of the voltage induced in a sensor
for controlled pressures varying in the range 1.5–4 kPa. We found a variation across
repetitions smaller than 2%, which corresponds to a very good accuracy.
Data Collection
The data from the six sensors were gathered simultaneously via a low-noise amplifier,
especially made for this research (sampling frequency: 4 kHz). Speech sounds were
recorded via a microphone with a multichannel audio recorder (M-Audio Fast Track Pro)
using Audacity software at 22.05 kHz. During the swallowing tasks the patient was
asked to manually push on a beep button, to signal his swallowing onset and offset.
The beep was loudly played in the room via a loudspeaker so that it could be recorded
with the microphone, with the same procedure as for the speech sounds. Pressure and
audio signals (speech and beeps) were a posteriori synchronized thanks to a square signal that was sent simultaneously to the low-noise
amplifier and to the audio recorder. The data acquisition was made by MATLAB software
and the data were displayed online as time-varying signals on the laptop screen.
During the recording, the subject was seated in a dental chair in an upright position
with the head stabilized by the headrest. For the speech production task described
in this paper, the subject was instructed to pronounce repetitions of two [i]C[a]
and two [a]C[i] nonsense words, where C was a French fricative consonant among /s/
and /z/ in different random order; /isa/ /asi/ /iza/ /azi/. The swallowing tasks consisted
of swallowing of saliva and 10 mL of water. Ten measurements were taken for each task.
Results
Examples of recorded data from the sensor located in the alveolar region of the PRESLA
system are shown in [Fig. 3], which show typical pressure patterns observed for the articulation of unvoiced
fricative /s/ in /isa/ (top panel), and for swallowing (bottom panel).
Fig. 3 Typical pressure signals measured with a sensor in the alveolar region of the PRESLA
system. Top panel: speech production, /s/ in /isa/; the vertical colored lines represent
acoustic time events; magenta = onset of consonantal noise, blue = offset of consonantal
noise, green = onset of postconsonantal vocal fold vibration, and yellow = onset of
postconsonantal laminar airflow. Bottom panel: swallowing, water swallow; the two
vertical lines mark the onset and offset times of swallowing as indicated by the subject
with the beeps. In both panels: Tons = onset time of positive pressure phase; Toff = offset time of positive pressure phase; Tmax = time of the maximum of the positive pressure phase. The green curve denotes the
first time-derivative of the pressure signal.
Relevant time events for the production of /s/, labeled using the spectrotemporal
characteristics of the speech signal, are indicated in the top panel with vertical
lines. We observe that the onset of mechanical pressure precedes the onset of the
consonantal noise. Hence, the noise does not start when tongue starts to be in contact
with the palate, but when the area of the vocal tract constriction is small enough
(below 30 mm according to Stevens,[4] p. 1188) to significantly increase the air pressure in the back cavity of the vocal
tract and generate turbulences in the constriction. Interestingly tongue pressure
further increases after the consonantal noise onset, reaches a maximum, and immediately
decreases after it. The offset of the consonantal noise occurs after the release of
the mechanical pressure against the palate, which shows that air pressure does not
instantaneously decrease.
For water swallow (bottom panel) the mechanical pressure reaches its maximum in the
very first part of the positive pressure phase, and regularly decreases in the second
half, which is consistent with the fact that swallowing starts with a sealing of the
vocal tract in its anterior part where the observed sensor is located. Interestingly,
the subject reported the onset of his swallow (first vertical line on the figure)
almost 200 ms after the onset of positive tongue-palate pressure in the alveolar region.
This suggests that the representation of the swallowing task for the subject could
be more associated with its pharyngeal phase than with its palatal phase.
To assess the general nature of the time variations depicted in [Fig. 3], we have provided for each task a statistical characterization of the pressure signal.
To do so, we have first time-normalized the pressure signals by dividing the time
by the duration of the positive pressure phase 
and we have then plotted the 95% confidence interval and the prediction interval
at one standard deviation of the time-normalized signals recorded for all the repetitions
of the same task. The 95% confidence interval corresponds at each normalized time
to the range of values in which the average value of the considered signal should
be with a 0.95 probability. The prediction interval refers for each normalized time
to the range of values that is likely to include the value of the measured signal
for any single new observation. Hence the prediction interval reflects at each normalized
time the uncertainty on the value of each single observation of the considered signal,
while the confidence interval reflects at each normalized time the uncertainty around
the mean value of the considered variable. [Figure 4] shows the results obtained for the repetitions of /s/ in /isa/ (Panel A) and water
swallow (Panel B). It can be observed that, consistent with the single observation
of [Fig. 3], pressure signals for /s/ are essentially bell shaped, with the decreasing phase
starting immediately after the end of the rising phase, whereas pressure signals observed
for water swallow begin with a maximum followed by a slow decrease during the rest
of the positive pressure phase. Moreover, we observed that the pressure magnitude
is almost 20 times larger for water swallow than for /s/ production. All these observations
are consistent with basic knowledge of the motor control mechanisms underlying each
task: in swallowing, tongue-palate contacts in the front part of the vocal tract must
be strong enough to provide a true sealing of the vocal tract and the tongue takes
advantage of these contacts to initiate its backward undulatory movement, which propels
the bolus toward the pharynx; in /s/ production, the tongue gently touches the palate
to stabilize the tongue and generate the small vocal tract constriction that gives
birth to air turbulences and to the consonantal noise.
Fig. 4 Statistical distributions of the tongue-palate mechanical pressure for fricative
/s/ in /isa/ (A) and water swallowing (B) for a sensor located in the alveolar region (front part of the complete denture).
Pressure time patterns normalized in time by the whole duration of the positive pressure
phase (see text for details). Dark gray shaded region: 95% confidence interval; light
gray shaded region: prediction interval at one standard deviation.
Discussion
Providing a precise description of the time variation of tongue-palate pressure in
speech production and swallowing should help formulate strong hypotheses about the
nature of motor control underlying the correct achievement of the tested motor tasks.
In a recently published paper,[3] we have shown that the production of the French velar stop /k/ is associated with
a plateau-like pattern of tongue pressure variation during vocal tract occlusion.
We have interpreted the plateau-like pattern as evidence for the hypothesis that the
production of velar stops is associated with tongue movement toward a virtual target
located above the palate, a target whose virtual achievement would be marked by the
pressure plateau. We have found above that the French fricative /s/ does not show
such a plateau-like pressure pattern, but rather a bell-shaped pattern with a significantly
smaller magnitude, suggesting a more gradual and less strong interaction between the
tongue and the palate. This could reflect a complex interplay between the tongue,
the palate, and the airflow going through the small constriction of the vocal tract
in the palate-alveolar region (see for example Perrier et al, 2000, for a modeling
of such an interplay).[5] We also observed a significantly different pressure pattern for water swallow, which
we could interpret in relation with the sealing of the vocal tract and the undulatory
movement of the tongue required in this task.
Our results confirm that the PRESLA system and its associated experimental protocol
involving complete denture wearers is well suited to the study of tongue–palate interactions
in speech production and swallowing. This can be done in old wearers of their prosthesis,
to study nonpathological condition, or with new wearers, to investigate adaptation
mechanisms. Obviously, further investigations are required to further pursue the issues
raised above. There are indeed obvious limitations in our study. First, data were
collected from a single subject. Inter-subject variability is a well-known characteristic
of oral motor functions, in particular speech production, and extending this experimental
protocol to a large number of subjects (classically 20–30 in speech production) is
a requirement, to draw general conclusions about speech and swallowing motor control
under normal conditions. Second, the calibration system of the sensors would beneficiate
from improvements that would enable pressure measurements over a range larger than
0–5 kPa, to validate our assumption of a linear voltage-pressure relation for the
whole domain of variation of the tongue-palate pressure. This is particularly important
for swallowing, in which pressure can be larger than 50 kPa. We are also working on
an improvement in the experimental protocol for swallowing. Currently, the postsynchronization
of the pressure signals with the actual motor task is based on the audio signal that
the subject has generated, when he was considering that he was swallowing. This induces
a temporal incertitude, due to the subjective evaluation by the subject of the onset
of his swallowing gesture and to the variability in the time interval between this
subjective onset and the actual hand gesture that pushes the beep button. Improvements
involve video recording of the neck region, to detect the onset of laryngeal elevation.
Conclusions
Using the temporal information of our recordings, patient can be taught to position
the tongue in purposeful manner to improve retention and stability of denture. Constant
repetition with use of coordinated exercises can help these patients learn muscular
activity patterns to aid in retention and stability.[6] For these reasons, the capability of our experimental device to record mechanical
interactions at multiple sites of prosthesis is an important step toward the specification
of general principles for the design of complete denture, to make most of the tongue
support and facilitate the good quality of speech and swallowing. These goals if achieved
will improve the comfort of edentulous patients.
Our device can also be useful together with other clinical evaluation to quantitatively
assess the impact of orofacial disorders such as myofascial pain,[7] maxillary odontogenic myxoma,[8] or associated with maxillofacial prosthetic obturator.[9]
