The present study aimed to verify the importance of postoperative articulatory rehabilitation in patients with oral cancer and to clarify the neurological changes underlying articulatory functional recovery. A longitudinal assessment of oral function and accompanying brain activity was performed using non-invasive functional magnetic resonance imaging (fMRI). We assessed 13 patients with cancers of the tongue and oral floor before and after ablative surgery. Articulatory function was assessed preoperatively and postoperatively using a conversation intelligibility test and the Assessment of Motor Speech for Dysarthria test. Patients also performed a verbal task during fMRI scans. The assessments were then repeated after the patients had undergone 4–6 months of articulatory rehabilitation therapy. Compared to pretreatment levels, articulatory rehabilitation resulted in a significant increase in activation in the supplementary motor cortex, thalamus, and cingulate cortex. The present study offers a quantitative assessment of the effects of speech rehabilitation by investigating changes in brain activation sites.
Malignant neoplasms, or cancers, are the leading cause of death worldwide. Although head and neck cancers account for only a small proportion of all cancers, the proportion of oral cancers among head and neck cancer cases is high, at 59%, and is increasing. Chemotherapy, radiotherapy, and surgery are the current treatments generally used for oral cancer.
Surgical treatment results in deformation or defects in the tongue, jaw, and surrounding tissues, which may lead to chewing and swallowing dysfunction and articulatory impairment, whereas functional impairments are arguably less severe following chemotherapy or radiotherapy. However, chemotherapy, radiation therapy, and chemoradiotherapy often result in adverse effects such as dysphagia and masticatory disorders due to mucositis and pharyngitis, as well as myelosuppression, nausea, vomiting, and functional impairments. Nevertheless, surgical treatment can also result in deglutition disorders, masticatory disorders, and dysarthria, similar to those observed after chemoradiotherapy, because it causes deformations and defects in the tongue, mandible, and surrounding tissues.
Oral functions are extremely important for dietary intake and swallowing, as well as for daily activities such as communication. For oral cancer patients, articulatory function deteriorates even if reconstructive surgery is performed following tumour resection. The accompanying decrease in conversation intelligibility has a great effect on the quality of life, regardless of age.
Postoperative articulatory impairment varies considerably depending on the extent, type, and size of the resected tumour. Regarding the advantages of articulatory rehabilitation, although several studies have been conducted using evaluations based on articulatory tests, only a few studies have used objective evaluations based on a clear understanding of how the extent of the resection and the differences in reconstruction methods have influenced the development of articulatory disorders after surgery. Extensive research has been conducted using articulation testing to assess the effectiveness of articulatory rehabilitation, although studies on the relationship between postoperative articulatory impairment and the extent of the resection and the method of reconstruction are limited. Because awareness is growing regarding the importance of rehabilitation in the field of oral surgery, additional reliable methods for the evaluation of articulatory function are needed in daily clinical practice; therefore, we conducted this research, focusing on functional magnetic resonance imaging (fMRI).
In recent years, neuroimaging techniques such as fMRI, magnetoencephalography (MEG), near-infrared spectroscopy (NIRS), and positron emission tomography (PET) have shown considerable advancements, and studies have utilized these techniques in the areas of rehabilitation and neuroscience. Research using fMRI to assess the recovery of brain function in patients with brain damage has been reported ; it has been shown that the recovery of motor function following brain damage caused by stroke is related to functional and structural reconstruction of the motor cortex and related areas based on brain plasticity. The usefulness of fMRI has been described extensively.
Haupage et al. used fMRI to assess the brain activation areas involved in postoperative changes in tongue movement and deglutition function in six tongue cancer patients. In these six patients, the brain activation regions at 6 months after surgery more closely resembled those of nine healthy individuals compared to the measurements taken in these six patients before surgery. Mosier et al. performed imaging analysis of tongue movements while swallowing saliva using fMRI in four tongue cancer patients who had been treated with partial tongue resection and primary plicature at 6 months after surgery; these data were subsequently compared with those obtained for eight healthy subjects. Their findings showed that strong activation of the parietal lobes and the cerebellum and changes in the brain cortex after partial tongue resection reflected adaptations in the biomechanics of tongue movements during swallowing. Their results indicated that, although the extent of resection affects tongue movement during deglutition, perception is unaffected. Both these studies compared healthy individuals with the postoperative conditions of tongue cancer patients, and the fMRI tasks included tongue tapping and saliva deglutition. Because the number of participants was small (i.e., n < 10), an individual assessment rather than a population-level analysis was conducted.
We believe that accurately assessing postoperative articulatory function in oral cancer patients requires evaluation of tongue movement as well as actual articulation, in addition to a comparison of the patient articulatory status before and after surgery and after articulatory rehabilitation. Conducting a population analysis may also be useful for verifying the relationship between brain activity and speech. Performing a population analysis of articulatory issues before surgery would greatly facilitate the identification of sites that are commonly activated in healthy subjects, and may play a major role in the creation of a database of fMRI experiments involving articulation. In addition, the creation of such a database may lead to new discoveries. Finally, performing a population analysis may identify common properties of the group, which would facilitate the characterization of individual differences within this group.
Thus, we employed non-invasive fMRI to examine brain activity during articulatory rehabilitation of oral cancer patients in order to establish the role of postoperative articulatory rehabilitation. The present study also aimed to improve and objectively assess the motivation of oral cancer patients to undergo articulatory rehabilitation based on patient feedback, which may facilitate the development of guidelines for rehabilitating postoperative oral cancer patients.
Materials and methods
The participants were recruited from patients who were examined and given a definitive diagnosis of tongue and oral floor cancer at our hospital. The patients included in this study required ablative surgery, had no history of psychiatric or neurological disease, and provided written consent after receiving a full explanation of the intent of the study. To avoid age- or sex-related biases in the data, we recruited patients of both sexes and across a broad range of ages. The study was approved by the institutional research ethics committee.
We examined 13 patients with cancer of the tongue and oral floor, including eight men and five women with a mean age of 65.8 years (range 41–82 years). The primary site was the tongue in nine cases and the oral floor in four. The resections were partial glossectomy ( n = 5), hemiglossectomy ( n = 3), subtotal glossectomy ( n = 1), and resection of the tongue and oral floor ( n = 4). The TNM classification was T1 in one case, T2 in six cases, T3 in three cases, and T4 in three cases. Disease was classified as stage I in one case, stage II in five cases, stage III in two cases, and stage IV in five cases ( Table 1 ).
|Case||Primary site||Resections||Reconstruction||TMN classification||Stage classification|
|1||Tongue||Partial glossectomy||No reconstruction||T2||Stage 2|
|2||Tongue||Partial glossectomy||No reconstruction||T1||Stage 1|
|3||Tongue||Partial glossectomy||No reconstruction||T2||Stage 2|
|4||Tongue||Partial glossectomy||No reconstruction||T2||Stage 2|
|5||Oral floor||Resection of the tongue and oral floor||Free ALT flap||T2||Stage 4|
|6||Tongue||Hemiglossectomy||Free ALT flap||T2||Stage 2|
|7||Tongue||Hemiglossectomy||Free ALT flap||T2||Stage 2|
|8||Tongue||Subtotal glossectomy||Free ALT flap||T3||Stage 3|
|9||Oral floor||Resection of the tongue and oral floor||Free RAMC flap||T4||Stage 4|
|10||Tongue||Partial glossectomy||Free FA flap||T4||Stage 4|
|11||Oral floor||Resection of the tongue and oral floor||Free RAMC flap||T3||Stage 3|
|12||Oral floor||Resection of the tongue and oral floor||Free ALT flap||T4||Stage 4|
|13||Tongue||Hemiglossectomy||Free ALT flap||T3||Stage 4|
The experimental procedures included a series of preoperative and postoperative examinations. Before surgery, each participant was examined in the rehabilitation department of our hospital, and articulation was assessed by a qualified speech therapist using the conversation intelligibility test and the Assessment of Motor Speech for Dysarthria (AMSD) test. The conversation intelligibility test ( Table 2 ) is used widely to evaluate articulatory function in oral cancer patients.
|Conversation intelligibility test|
|3||Intelligible when the topic is known|
Conversation intelligibility, which refers to the ease with which a person’s speech can be understood, is the most important indicator of communicative competence in oral communication. Kawaguchi et al. evaluated the communication skills of 59 postoperative lingual cancer patients based on intelligibility and showed that patients with speech clarity of 80% or higher had no trouble in their daily conversations, whereas those with speech clarity of 40% or lower experienced difficulties in their daily conversations. The test is an effective method of objectively evaluating the severity of articulatory impairment in oral and oropharyngeal cancer patients. Postoperative speech status was also assessed using the conversation intelligibility test, which has been used widely in Japan to assess speech impairment in postoperative oral cancer patients.
The AMSD ( Tables 3 and 4 ) is a test that evaluates speech impairments resulting from motor dysfunction of the vocal apparatus arising from nervous or muscular system lesions. It enables evaluation of the physiological functions of each organ related to articulation from the perspective of motor function.
|Naturalness of conversation test|
|Articulation characteristics rating|
|0||Excellent (no altered articulation characteristics)|
|1||Slight disorder (a few altered articulation characteristics)|
|2||Moderate disorder (clearly altered articulation characteristics)|
|3||Severe disorder (many altered articulation characteristics)|
We conducted these two tests to improve the accuracy of the results for articulatory function from articulation tests. Additionally, fMRI was used to evaluate preoperative articulatory function before tumour resection. After confirming postoperative recovery of articulation and swallowing ability, fMRI was performed with the same task as before surgery, and a postoperative assessment was made by a speech therapist. Articulatory rehabilitation was initiated during postoperative hospitalization under the guidance of a speech therapist, and rehabilitation was continued for 4–6 months on an outpatient basis after the patient was discharged. At the completion of articulatory rehabilitation, articulatory function was assessed with fMRI and speech therapy evaluation. Two speech therapists with five or more years of experience assessed the preoperative and postoperative speech conditions of the patients and designed therapies as needed. Speech therapies were centred on articulation training, deglutition training, and neck movement.
The principle of fMRI is based on the blood oxygenation level dependency (BOLD) effect, which detects signal fluctuations due to changes in the ratio of oxyhemoglobin and deoxyhemoglobin in the blood in relation to brain activity. fMRI images were captured with ultra-fast echo planar imaging (EPI). T1-weighted images were also collected to check the activation sites based on their anatomical position.
MRI equipment and data analysis – Statistical Parametric Mapping
Statistical Parametric Mapping v. 5 (SPM5; Wellcome Department of Imaging Neuroscience, University College, London, UK ) is a software programme that was created in accordance with international standards for conducting imaging analysis of brain function.
The MRI scanner used in this study was the Intera Achieva 1.5T (Koninklijke Philips Electronics N.V., Eindhoven, Netherlands), and images were taken using gradient-recalled EPI. The imaging conditions were repetition (TR) = 3000 ms, echo time (TE) = 40 ms, flip angle (FA) = 90°, matrix = 64 × 64, 30 slices, voxel size = 3.59 × 3.65 × 4.00 mm, slice thickness = 4.0 mm, and slice gap = 0.4 mm. In one session, 120 images were taken; the first eight were excluded and the remaining 112 were analyzed. To confirm the anatomical location of the activation sites, three-dimensional images of the brain were constructed with T1-weighted images, which provide a high-resolution display of anatomical structures (i.e., shape) by using a method that accentuates the fat content in the body. When making the EPI and T1-weighted images, a cushion or head belt was used to stabilize the head to minimize artefacts from head movement.
The experimental task was to vocalize short sentences during the MRI scan. The task had a block design in which a presented short sentence was repeated for 21 s followed by a 21-s rest as a single cycle. The MRI device that we used in this experiment had a maximum EPI imaging time of 3 min; therefore, to perform an eight-block task and an eight-block rest repeatedly, one block experiment lasted for 21 s. This was repeated for a total of eight cycles.
The participants were provided headphones and were instructed to listen to short sentences spoken by an experimenter and then to repeat these sentences several times. During the rest periods, they were instructed to remain still and silent. The instructions for the experiment were given remotely from a control room. The sentences used in the articulatory rehabilitation were used as the short sentences to be pronounced ( Table 5 ).
|Pronounced short sentences|
|1||Japanese sentence: sa.sa.yaku.yoona.asase.no.seseragi
English meaning: A little stream of a shallow is a whisper
|2||Japanese sentence: ruri.mo.harimo.terase.ba.hikaru
English meaning: You can’t teach an old dog new tricks
|3||Japanese sentence: takai.takai.tokoro.ni.noboru.tokoro.da
English meaning: I have reached a very high place
|4||Japanese sentence: karada.ga.daruku.te.daruku.te.shikata.ga.nai
English meaning: I feel tired and very lethargic
The data obtained were analyzed with SPM5 brain function image analysis software operated in a MATLAB. Because brain activity differs among individuals, the same activated sites were examined in each participant, and a group analysis with the 13 patients was performed to detect trends. The uncorrected threshold value in the group analysis was P < 0.001.
To correct for head movement produced by body movement, respiration, or heartbeat during EPI, image data fluctuations accompanying head movement were corrected by combining initial brain imaging data with subsequent brain imaging data. In the current experiment, we confirmed that parallel displacement did not exceed 2 mm and rotation did not exceed 1° on the x -axis (left–right), y -axis (antero-posterior), or z -axis (cranio-caudal).
Two different types of MRI image were utilized for brain function analysis using fMRI. The first was an EPI image, on which analyses have been conducted thus far. However, because the low resolution of EPI images prevents their use in the accurate identification of active sites, other high-resolution anatomical images must be obtained and subsequently aligned with the EPI images.
We performed displacements, rotations, zooming, and transformations to correct misalignments between the EPI images and the anatomical images due to differences in imaging parameters, and we aligned the images by adjusting their positions.
Smoothing is a process that consists of smoothing images to correct for the anatomical variability among subjects and to facilitate statistical data processing.
Activity models were prepared under each condition for each individual, and individual and group analyses were conducted. Regions of interest were analyzed to investigate the relative activity during task performance and at rest, as well as the activity in specific brain regions during articulation. The WFU_PickAtlas (ANSIR Laboratory; Wake Forest University School of Medicine, Winston-Salem, NC, USA) was used to anatomically identify activation regions. For sites with significant changes in BOLD signals, we investigated the cerebral cortex and activation sites, the regions based on Brodmann’s areas (BA), the number of voxels within each cluster, the Z -value of activation intensity at each active site, and the positions indicating the x , y , and z coordinates in the standard brain. The coordinates obtained with SPM5 were Montreal Neurological Institute (MNI) coordinates; these were converted to Talairach coordinates in order to compare them with other studies, and they are expressed in this form in the text. The brain activity sites were identified. The same activated sites were examined in each participant because brain activity differs among individuals, and a group analysis was performed with the 13 patients to detect trends. The uncorrected threshold value in the group analysis was P < 0.001.
Because brain activity differs among individuals, the same activated sites were extracted for each participant, and a group analysis with the 13 patients was performed to detect trends.
Regions that showed significant increases in brain activity upon articulation before surgery ( Fig. 1 A ), after surgery ( Fig. 1 B), and after articulatory rehabilitation ( Fig. 1 C) are shown in sagittal, coronal, and axial images, superimposed with the reconstructed two-dimensional images of the brain surface, in Fig. 1 . In addition, we calculated the x , y , and z coordinates corresponding to the Talairach coordinates, the cerebral cortical and internal activation sites, the BAs, the size of each cluster, and the Z -value that represents the intensity of brain activation during articulation before surgery, after surgery, and after articulatory rehabilitation ( Table 6 ).