Analysis of using the tongue deviation angle as a warning sign of a stroke
BioMedical Engineering OnLine
Analysis of using the tongue deviation angle as a warning sign of a stroke
Ching-Chuan Wei 0 1
Shu-Wen Huang 3
Sheng-Lin Hsu 0
Hsing-Chung Chen 2
Jong-Shin Chen 0
Hsinying Liang 0
0 Department of Information and Communication Engineering, Chaoyang University of Technology , Taichung , Taiwan
1 168 , Jifeng E. Rd., Wufeng District, Taichung, 41349 Taiwan
2 Department of Computer Science and Information Engineering, Asia University , Taichung , Taiwan
3 Department of Health, Executive Yuan, Graduate Institute of Informatics, Chaoyang University of Technology and Taichung Hospital , Taichung , Taiwan
Background: The symptom of tongue deviation is observed in a stroke or transient ischemic attack. Nevertheless, there is much room for the interpretation of the tongue deviation test. The crucial factor is the lack of an effective quantification method of tongue deviation. If we can quantify the features of the tongue deviation and scientifically verify the relationship between the deviation angle and a stroke, the information provided by the tongue will be helpful in recognizing a warning of a stroke. Methods: In this study, a quantification method of the tongue deviation angle was proposed for the first time to characterize stroke patients. We captured the tongue images of stroke patients (15 males and 10 females, ranging between 55 and 82 years of age); transient ischemic attack (TIA) patients (16 males and 9 females, ranging between 53 and 79 years of age); and normal subjects (14 males and 11 females, ranging between 52 and 80 years of age) to analyze whether the method is effective. In addition, we used the receiver operating characteristic curve (ROC) for the sensitivity analysis, and determined the threshold value of the tongue deviation angle for the warning sign of a stroke. Results: The means and standard deviations of the tongue deviation angles of the stroke, TIA, and normal groups were: 6.9 3.1, 4.9 2.1 and 1.4 0.8 degrees, respectively. Analyzed by the unpaired Student's t-test, the p-value between the stroke group and the TIA group was 0.015 (>0.01), indicating no significant difference in the tongue deviation angle. The p-values between the stroke group and the normal group, as well as between the TIA group and the normal group were both less than 0.01. These results show the significant differences in the tongue deviation angle between the patient groups (stroke and TIA patients) and the normal group. These results also imply that the tongue deviation angle can effectively identify the patient group (stroke and TIA patients) and the normal group. With respect to the visual examination, 40% and 32% of stroke patients, 24% and 16% of TIA patients, and 4% and 0% of normal subjects were found to have tongue deviations when physicians A and B examined them. The variation showed the essentiality of the quantification method in a clinical setting. In the receiver operating characteristic curve (ROC), the Area Under Curve (AUC, = 0.96) indicates good discrimination. The tongue deviation angle more than the optimum threshold value (= 3.2) predicts a risk of stroke. (Continued on next page)
Stroke, a cerebral vascular incident, is mainly caused by abnormal blood vessels in the
brain. According to the statistical results of the World Health Organization (WHO),
stroke remains the worldwide second leading cause of death. It is estimated that one in
five stroke survivors will have the chance of a second stroke within five years. Thus, it
has a high recurrence rate, and recurrence can bring about disability and dementia,
often leading to a heavy burden for an individual household, a community, and
ultimately society in general. This reminds us how important it is to prevent, recognize and
monitor the stroke subject.
A stroke, often occurring suddenly, happens for two main reasons. Firstly, a
hemorrhagic stroke results from a weakened vessel that ruptures and bleeds into the
surrounding brain. About 20% of strokes are hemorrhagic. Secondly, and more
commonly, an ischemic stroke occurs when an artery in the brain becomes blocked. About
80% of strokes are ischemic strokes. If a major artery to the brain is blocked, part of
the brain tissue can die from lack of oxygen carried in the blood. When brain tissue
dies, the effects on the body will depend on which body functions that part of the brain
controls. Some of the effects on the body are quite well known and are commonly
recognized as the result of a stroke. A stroke will produce changes in the body and
affect various functions, including the sensory function, action function, language
ability, the swallowing function, etc. For instance, paralysis of the right side of the body
(right-sided stroke) will be caused by damage to the left half of the brain. In most
patients, inability to speak will also be due to damage to the left side of the brain,
because the left side of the brain controls speech.
A mini-stroke, also called a transient ischemic attack (TIA), causes a reversible
neurologic deficit. The damaged area can be so small that the blood supply to other
parts of the brain can compensate, and a full recovery takes place. Dizziness and
giddiness, blurred vision, and unsteadiness are the typical symptoms of a TIA. These
symptoms resolve themselves in less than 24 hours. People often disregard the mini-stroke,
failing to recognize that the TIA may be a warning sign that a more severe stroke may
take place . It is not uncommon for those who experience but disregard the TIA
to have a stroke several days later. Therefore, recognizing the symptoms of a TIA and
recognizing the TIA as a warning may make it possible to prevent a stroke.
The tongue in mammals has important motor and sensory functions. When the
motor cortex in the brain is damaged, the hypoglossal nerve, which is a pure motor
nerve innervating the muscles of the tongue, will be defective. Therefore, the tongue
will have a tendency to turn away from the midline when extended or protruded, and it
will deviate toward the side of the lesion. This is called tongue deviation [2-5]. Hence,
the symptom of tongue deviation is observed in a stroke or TIA [5-8]. For thousands of
years, the deviation of the tongue has also been recognized as a symptom of what is
called a wind stroke in traditional Chinese medicine [9-11]. The symptom of tongue
deviation in stroke patients has been observed from ancient to modern times. Many
people may recognize the more common symptoms of a stroke, such as slurred speech
or paralysis of one side of the body; however, fewer are familiar with tongue deviation.
Nevertheless, there is much room for the interpretation of the tongue deviation test.
How crooked is crooked? How far to one side does the tongue have to be regarded as a
clear sign of a stroke having occurred? There are too many variables with tongue
deviation. The crucial factor is the lack of an effective quantification method of tongue
deviation. If we can quantify the features of the tongue deviation and scientifically
verify the relationship between the deviation angle and a stroke, the information provided
by the tongue will be helpful in recognizing a warning sign of a stroke. In this study,
we developed a simple and effective method to quantify the deviation angle of the
crooked tongue, and we conducted the experiments to verify the feasibility of using the
tongue deviation angle as the warning sign of a stroke. Finally, we used the receiver
operating characteristic curve (ROC) for the sensitivity analysis, and determined the
threshold value of the tongue deviation angle for the warning sign of a stroke.
Tongue image acquisition and edge segmentation
We built a brace rack to support the chin in order to fix the tongue position. Subjects
were asked to make the tongue protrude. Then, we took a picture including the
subjects tongue using a digital camera and a circular light source, which uniformly
distributes the light on the tongue, as shown in Figure 1. Next, we adopted the threshold
method using Otsus thresholding algorithm and filtering process for edge
segmentation because they can achieve an easy, fast, and effective segmentation result with the
tongue image . Thus, we were able to remove skin, tooth, lip, background, etc., to
obtain the pure tongue image, as shown in Figure 2 [12-15]. In the following steps,
we will start to quantify the angle of tongue deviation.
Figure 1 The captured tongue image.
Figure 2 The pure tongue image through edge segmentation.
1. The searching order of the left oblique starts from point (0,0), and then (0,1), (1,0), (0,2), (1,1), (2,0), etc., until the first point encountering the red tongue is found. The first point is called the left oblique point.
2. The searching order of the right oblique starts from the top right point (m, n),
and then (m-1, n), (m, n-1), (m-2, n), etc., until the first point encountering the red
tongue is found. The first point is called the right oblique point. The middle point
between the left and right oblique points is defined as the root point of the tongue.
It is shown as a star point in Figure 3(b) and is marked as symbol A.
1. The vertical coordinate of the tongue center point: The top point and bottom
point in Figure 4 are the highest and lowest points of the tongue, respectively. The
value of the vertical coordinate of the tongue center point is determined by the
average value of the vertical coordinate of the top and bottom points in the tongue,
as shown in Figure 4.
2. The horizontal coordinate of the tongue center point: The left point and right
point in Figure 4 are defined by the intersection points between the tongue edge and
the horizontal line, which crosses the middle point between the top and bottom
points. The horizontal position of the tongue center point was thus derived by the
horizontal value of the middle point between the left and right points in the tongue,
as shown in Figure 4. As soon as the horizontal position of the tongue center point
was found, the center point of the tongue was specifically determined. It was shown as
a star point in Figure 4 and marked as symbol B. The tongue center point means
the center of the tongue; thus, it may deviate with an angle compared to the vertical
line in stroke patients, as shown in Figure 5.
Evaluating the tongue deviation angle
The value of the vertical coordinate of point C in Figure 5 is equivalent to that of point
B, and the value of the horizontal coordinate of point C is equivalent to that of point
A. Thus, point C is defined, and C in Figure 5 forms a right angle. A in Figure 5 is
the tongue deviation angle used to recognize stroke patients. A is then calculated by
Reproducibility of the tongue deviation angle
Both operators were given a brief practical introduction to the technique of the tongue
deviation angle, and then they performed 30 practice measurements over 5 days before
Figure 4 The star marks the center of the tongue image.
Figure 5 The tongue image of a stroke patient, and the related deviation angle.
embarking on the experiments. After taking the first capture of the tongue image by
operator 1 and a 10-minute rest, we conducted the second capture by operator 2 in
order to assess the reproducibility of the tongue deviation angles between operators.
Next, after a 10-minute rest, we conducted the third capture again by operator 1 to
assess the within-operator difference. There were 25 subjects involved in the
reproducibility experiments. The mean and standard deviations of the capture on tongue deviation
angles are listed in Table 1.
SPSS software (SPSS 12.0, SPSS Inc., USA) was used to compute the Intra-class
correlation coefficient (ICC). ICCs between the first and second measurement, and
between the first and third measurement, are 0.85 and 0.89, respectively. The
aforementioned results imply the excellent reliability for the intra-rater and inter-rater
analyses. Consequently, there is high reproducibility in the measurements of the tongue
The subjects enrolled in the experiment included three groups. The first group
comprised 25 stroke patients undergoing treatment, ranging between 55 and 82 years of
age (15 males and 10 females). The second group comprised 25 TIA patients
undergoing treatment, ranging between 53 and 79 years of age (16 males and 9 females), and
the third group comprised 25 normal subjects with no stroke or TIA, ranging between
52 and 80 years of age (14 males and 11 females). The age distribution of the three
Table 1 Analysis of reproducibility of the tongue deviation angle
Variable Mean (degree)
1st measurement (operator 1) 1.63
2nd measurement (operator 2) 1.57
3rd measurement (operator 1) 1.48
groups had no statistical differences (p < 0.001). None of the subjects in the stroke
group and the TIA group overlapped.
The members of the three groups had tongue images captured from 9/10/2011 to 3/
05/2012. During this time, the subjects were visually inspected by two physicians
(named A and B) using the same standard to identify whether the subjects had
tongue deviations. The tongue was observed from the front with the examiner
manually correcting any confounding mouth asymmetry. The position and configuration of
the median raphe in relation to the nasal bridge were used to estimate the deviation.
All participants were asked not to imbibe any alcoholic or caffeinated beverages on the
day of the experiment. The experiment protocol was approved; written informed
consent was obtained from all of the participants before they enrolled in this study.
Figure 6(a), 6(b) and 6(c) show the tongue images of 3 typical normal subjects,
respectively with the tongue deviation angles of 0.74, 1.2, and 2.5 degrees. Figure 7(a),
7(b) and 7(c) show the tongue images of 3 typical TIA subjects, respectively with the
tongue deviation angles of 4.3, 6.3, and 6.6 degrees. Figure 8(a), 8(b) and 8(c) show the
tongue images of 3 typical stroke patients, respectively with the tongue deviation angles
of 8.1, 10.8 and 19.5 degrees. Table 2 shows a summary of the means and standard
deviations for the clinical characteristics of the three groups, including age, weight,
height, BMI (Body Mass Index), and tongue deviation angle.
The means and standard deviations of the stroke, TIA, and normal groups were
6.9 3.1, 4.9 2.0 and 1.4 0.8 degrees, respectively. Figure 9 shows the box plot of the
distribution of the stroke, TIA, and normal groups. We used the Shapiro-Wilk test in
SPSS software to verify the normality of the stroke, TIA, and normal groups. All the
corresponding p-values are greater than 0.05; thus, the distributions of the stroke, TIA,
and normal groups are normal, respectively. Next, analyzed by unpaired Students
ttest, the p-value between the stroke and TIA groups was 0.015 (>0.01), indicating no
significant difference in the tongue deviation angle. The p-values between the stroke
and normal groups, as well as between the TIA and normal groups, were both less than
0.01. These results show the significant differences in the tongue deviation angle
between the patient groups (stroke and TIA patients) and the normal group. These
results also imply that the tongue deviation angle can effectively identify the patient
groups (stroke and TIA patients) and the normal group. With respect to the visual
examination of physician A, 40% of stroke patients, 24% of TIA patients, and 4% of
normal subjects were found to have tongue deviations. As for the visual examination
by physician B, 32% of stroke patients, 16% of TIA patients, and 0% of normal
subjects were found to have tongue deviations [5,16].
25 TIA patients and 25 normal subjects were included in the sensitivity analysis. We
used the receiver operating characteristic curve (ROC) shown in Figure 10 to assess the
sensitivity of the proposed method to evaluate the occurrence of a stroke. The Area
Under Curve (AUC) (= 0.96) indicates good discrimination. The optimum operating
point of the ROC curve corresponds to the threshold value of the tongue deviation
angle of 3.2. Thus, a tongue deviation angle greater than 3.2 may preliminarily predict
a risk of a stroke.
Various techniques may assist in monitoring the brain. Examples include continuous
transcranial Doppler, near-infrared spectroscopy, measurement of tissue oxygen,
somatosensory evoked potentials and continuous electroencephalography (cEEG). For
example, previous studies have shown that EEG provides physiological information
following a stroke in the lesion site relating to the aspects of location, grade of damage,
and physiological recovery [17-19]. Quantitative electroencephalography (qEEG)
measures of delta (1-4 Hz) power and delta/alpha ratio (DAR) have been shown to be
relevant to the site of ischemic lesion and effective in predicting the recovery from a stroke
as well [17,20,21]. In addition, the brain symmetry index (BSI) was proposed to quantify
the ischemic damage, which was also related to clinical acute ischemic hemispheric
stroke [18,22]. Furthermore, the effectiveness of nonlinear analysis and detrended
fluctuation analysis (DFA) was also reported in the case of subcortical strokes [17,23].
The tongue is a sensitive area of the human body as a result of its extensive neural
controls. Consequently, even minor hypoglossal nerve damage resulting from a stroke
may have a significant impact on the shape and motion of the tongue. Thus, the tongue
deviation angle may be an indicator for determining the health or balance status of the
nervous system before we resort to using expensive equipment. The experimental
results indicated that almost all the participants, including stroke patients, TIA
patients, and normal subjects had a nonzero deviation angle. The major differences
existed in the degree of deviation, i.e., serious, moderate, or light deviation. Generally
speaking, the stroke patients had larger average deviation angles (mean = 6.9 degrees),
the TIA patients had smaller average deviation angles (mean = 4.9 degrees), and normal
subjects had the smallest average deviation angles (mean = 1.4 degrees). Consequently,
the clinical physician should not identify stroke patients based on the presence or
absence of tongue deviation; instead, the physician should evaluate the degree of the
tongue deviation angle.
With regard to the experimental results of visual inspection, there was a variation to
some extent between physicians A and B. This is not surprising since visual illusion
is a common occurrence for most people. A visual illusion is characterized by visually
perceived images that differ from objective reality. The information gathered by the
eye is processed in the brain to give a perception that does not conform to a physical
measurement of the stimulus source [24-26]. For example, Figure 7(a), 7(b), 7(c), 8(a)
Figure 9 The box plot of the distribution of the stroke, TIA, and normal groups.
Figure 10 The receiver operating characteristic curve.
and 8(b) present the images of 5 patients respective tongues with the deviation angles
of 4.3, 6.3, 6.6, 8.1 and 10.8 degrees. We compare these with all the normal subjects,
whose deviation angles show a distribution of 0 ~ 3 degrees. Even for such an obvious
angle distinction in numeric values between the patients and the normal group, it is
hard to determine by visual examination that the 5 patients have tongue deviations.
Hence, these cases were diagnosed as no tongue deviation by both physicians A and
B. The main reason for this may be that these angles are beyond the ability of the
ordinary eye to distinguish from surrounding objects.
According to previous research, the percentage of stroke patients with tongue
deviation is about 29% . Due to visual illusion, the percentage is not high enough to
make a case for allowing tongue deviation be a dependable indicator of stroke.
Therefore, because relying on the eyes to determine the measurement of the deviation angle
allows too much room for error, a quantification method is essential for clinical
application. In this study, we developed a quantification method of tongue deviation angle
and verified its effectiveness. We also found that even normal subjects have nonzero
tongue deviation angles. The major differentiating factor between normal subjects and
stroke patients is the degree of the tongue deviation angle. Using the tongue deviation
angle of 3.2 discussed above as the threshold value, 84% of TIA patients and 88% of
stroke patients are considered to have obvious tongue deviations. The percentages are
much higher than that in reference 5, and they are high enough to be an indicator
A stroke is often referred to by doctors as a cerebrovascular accident, but a stroke is
rarely an accident. The underlying conditions of a stroke are usually present for years
before a stroke occurs, although the symptoms of a stroke may occur suddenly. The
mean value of the tongue deviation angles of the TIA patients is less than that of the
stroke patients, which explains the above statement. TIAs have the same symptoms as
a stroke, but they are temporary and do not usually cause long-term brain damage. Just
like full strokes, TIAs need emergency treatment and should not be ignored. A person
who has had a TIA is at greater risk of having a stroke. Thus, a TIA, or mini-stroke, is
a warning of an impending stroke. In other words, if we can identify the occurrence of
TIA using the tongue deviation angle, we can present the warning sign of a stroke. By
recognizing the warning sign and taking action, the patients may be able to prevent a
stroke or reduce its severity because they are able to obtain medical help quickly.
Our results indicated that this quantification method can effectively distinguish
between the patient groups (stroke and TIA patients) and the normal group. In
particular, the method can distinguish between the TIA patients and the normal subjects.
Because a TIA is a warning sign in the early stage of a stroke, this result demonstrated
the feasibility of using the proposed quantification method as a prediction factor of a
stroke. Capturing the tongue images of the patients at high risk of stroke, we may find
the patients with larger tongue deviation angles in advance. In addition, if capturing the
tongue image in a routine health examination, we may also identify the subjects with
larger tongue deviation angles. Therefore, the presence of tongue deviation angles
greater than 3.2 should alert the clinician to take the necessary precautions to avoid
more dangerous complications for stroke patients. Preventing the complications of
stroke will help to reduce a heavy burden on families and society. This quantification
method of the tongue deviation angle is proposed for the first time for recognizing
stroke patients. It improves the clinical tongue diagnosis of stroke. The experiments
comparing normal and patient groups showed the methods effectiveness. Due to the
methods simplicity, it may be applicable in future homecare or telehealth to present
the early warning of the occurrence of a stroke or to show the prognosis.
CCW carried out the experiment design, analysis, and wrote the paper; SWH and SLH contributed to the collection
and analysis of the tongue images; HCC, JSC and HL worked on the image processing of tongue segmentation and
analysis. All authors read and approved the final manuscript.
1. Rothwell PM , Johnston SC : Transient ischemic attacks: stratifying risk . Stroke 2006 , 37 : 320 - 322 .
2. Wiebers DO , Feigin VL , Brown RD : Handbook of Stroke . Philadelphia: Lippincott Williams & Wilkins; 2006 .
3. Geyer JD , Gomez CR : Stroke: a Practical Approach . Philadelphia: Lippincott Williams & Wilkins ; 2009 : 346 - 348 .
4. Urban PP , Hopf HC , Connemann B , Hundemer HP , Koehler J : The course of cortico-hypoglossal projections in the human brainstem: functional testing using transcranial magnetic stimulation . Brain 1996 , 119 : 1031 - 1038 .
5. Umapathi T , Venketasubramanian N , Leck KJ , Tan CB , Lee WL , Tjia H : Tongue deviation in acute ischaemic stroke: a study of supranuclear twelfth cranial nerve palsy in 300 stroke patients . Cerebrovasc Dis 2000 , 10 : 462 - 465 .
6. Morgan JC , Alhatou M , Oberlies J , Johnston KC : Transient ischemic attack and stroke associated with sildenafil (Viagra) use . Neurology 2001 , 57 : 1730 - 1731 .
7. Greenstein AJ , Chassin MR , Wang J , Rockman CB , Riles TS , Tuhrim S , Halm EA : Association between minor and major surgical complications after carotid endarterectomy: results of the New York carotid artery surgery study . J Vasc Surg 2007 , 46 : 1138 - 1146 .
8. Yoon SS , Park KC : Glossoplegia in a small cortical infarction . J Neurol Neurosurg Psychiatry 2007 , 78 : 1372 .
9. Wiseman N , Ellis A : Fundamentals Chinese Medicine. Massachusetts USA: Paradigm Publications ; 1996 : 96 - 97 .
10. Chen YM , Yang ZZ , Luo CY : TCM Tongue Inspection Primer . Shantou: Shantou University Press ; 2006 .
11. Maciocia G : Tongue Diagnosis in Chinese Medicine . Seattle USA: Eastland Press ; 2000 .
12. Wang CH , Wei CC , Li CH : Comparison of edge segmentation methods to tongue diagnosis in traditional Chinese Medicine . Lecture Notes in Comput Sci 2010 , 6422 : 232 - 238 .
13. Li W , Hu S , Wang S , Xu S : Towards the objectification of tongue diagnosis automatic segmentation of tongue image . 35th Annual Conference of IEEE Industrial Electronics ; 2009 : 2121 - 2124 .
14. Pang B , Zhang D , Wang K : The bi-elliptical deformable contour and its application to automated tongue segmentation in Chinese medicine . IEEE Trans Med Imaging 2005 , 4 : 946 - 956 .
15. Zhu WN , Zhou CL , Xu D , Xu J : A multi-feature CBIR method using in the traditional Chinese medicine tongue diagnosis . IEEE 1st Int Symp on Pervasive Comput and Appl 2006 , pages: 831 - 836 .
16. Venkatesubramanian N , Sadasivan B , Tan AKY : Stroke pattern in a Singapore hospital-based stroke data bank . Cerebrovasc Dis 1994 , 4 : 250 .
17. Molnar M , Csuhaj R , Horvath S , Vastagh I , Gaal ZA , Czigler B , Balint A , Csikos D , Nagy Z : Spectral and complexity features of the EEG changed by visual input in a case of subcortical stroke compared to healthy controls . Clin Neurophysiol 2006 , 117 : 771 - 780 .
18. van Putten MJAM , Tavy DLJ : Continuous quantitative EEG monitoring in hemispheric stroke patients using the brain symmetry index . Stroke 2004 , 35 : 2489 - 2492 .
19. Van Huffelen AC , Poortvliet DC , Van der Wulp CJ : Quantitative electroencephalography in cerebral ischemia . Detection of abnormalities in normal EEGs. Prog Brain Res 1984 , 62 : 3 - 28 .
20. Finnigan SP , Walsh M , Rose SE , Chalk JB : Quantitative EEG indices of sub-acute ischaemic stroke correlate with clinical outcomes . Clin Neurophysiol 2007 , 118 : 2525 - 2532 .
21. Finnigan SP , Rose SE , Chalk JB : Contralateral hemisphere delta EEG in acute stroke precedes worsening of symptoms and death . Clin Neurophysiol 2008 , 119 : 1690 - 1694 .
22. van Putten MJAM : Extended BSI for continuous EEG monitoring in carotid endarterectomy . Clin Neurophysiol 2006 , 117 : 2661 - 2666 .
23. Hwa RC , Ferree TC : Stroke detection based on the scaling properties of human EEG . Physica A: Stat Mech and Its Appl 2004 , 338 : 246 - 254 .
24. Eagleman DM : Visual Illusions and Neurobiology . Nat Rev Neurosci 2001 , 2 : 920 - 926 .
25. Solso RL : Cognitive psychology. 6th edition . Boston: Allyn and Bacon ; 2001 .
26. Jung YM , Shen J : First-order modeling and stability analysis of illusory contours . J Visual Comm Image Representation 2008 , 19 : 42 - 55 .