27th National Clinical Education Symposium Presentation Abstracts
30 APRIL 2025, WEDNESDAY
13:00-14:00 ORAL PRESENTATION SESSION - 13
Development Study of Turkish Speech Analysis for Major Neurocognitive Disorder Due to Alzheimer’s Disease
H. Mihrimah Öztürk1, Saadin Oyucu2, Hüseyin Polat3, Özgür Aydın4, Erguvan Tuğba Özel Kızıl5
-
Show more (Affiliations)
1. Faculty of Medicine, Kirikkale University, Kirikkale, Türkiye
2. Faculty of Engineering, Adiyaman University, Adiyaman, Türkiye
3. Faculty of Technology, Gazi University, Ankara, Türkiye
4. Faculty of Language and History-Geography, Ankara University, Ankara, Türkiye
5. Faculty of Medicine, Ankara University, Ankara, Türkiye
DOI: 10.5080/kes27.abs148 Page 1-3
BACKGROUND AND AIM: Alzheimer’s disease (AD) is the leading cause of dementia/major neurocognitive disorder in the elderly. It is a neurodegenerative disease characterized by progressive loss of cognitive functions, and short-term memory impairment is considered as the core symptom. However, language skills are also affected in AD (Ivanova,2024). Early diagnosis and management of Alzheimer’s disease are very important for the quality of life of both patients and caregivers (Sanz et.al.,2022). Although screening tests used for neuropsychological assessment help detect cognitive deficits at an early stage, their sensitivity is limited, and they should be applied by trained specialists. In this respect, speech and language disorders seem to be an important tool for early diagnosis of AD. However, in recent years, advances in the fields of classification, voice processing and speech to text have made it possible to diagnose diseases even more easily (Vigo et.al,2022). The diagnosis of AD based on linguistic features and speech is a relatively new field and so far, from a computational/ algorithmic perspective, there is no established and widely accepted method. There are a limited number and quality of studies conducted in Turkish. Therefore, the results obtained from this study, in which 108,169 seconds of speech recordings of a total of 105 participants were analyzed, in which AH-MND, MİND and healthy elderly were evaluated together with detailed clinical, neuropsychological evaluation and language analysis, are important.
METHODS: The aim of this study was to investigate whether it is possible to distinguish individuals with major neurocognitive disorder due to Alzheimer Disease (AH-MND) (n=41) or minor neurocognitive disorder (MiND) (n=29) from healthy elderly (n=35) using speech analysis. In order to evaluate cognitive functions, Standardized Mini Mental Test, Clock Drawing Test, Montreal Cognitive Assessment Scale, Öktem’s Auditory Verbal Learning Test, Verbal Fluency Test, Augmented Cued Recall Test and Trail Making Test were applied to all participants. In order to evaluate language functions, the participants were administered the Cookie Theft Picture Description Test, which is a part of the Boston Aphasia Test. In addition, for the evaluation of spontaneous speech, three questions such as ‘Can you tell me about your ordinary day?’, ‘Can you tell me about your happiest moment?’, ‘Can you tell me about your unhappiest moment?’ were asked to the participants and the participants were asked to tell their memories in a logical order of events. In this test, it was ensured that the participants felt more comfortable without any visual stimuli and without any restrictions, and it was aimed to reveal their verbal expression skills more clearly. During the application of the tests, one-to-one communication was established with the participants and audio recordings were taken during the Verbal Fluency tests, Cookie Theft Picture Description Test, Boston Naming Test and spontaneous speech evaluations. The recordings were analyzed in terms of prosodic, lexical and acoustic properties using automatic speech recognition and PRAAT applications. The recorded speech samples were transcribed both manually and with the help of Automatic Speech Recognition (ASR) system. The transcripts of the participants’ speech were used to identify the most frequently used linguistic measures in the literature that showed statistically significant differences (Word count, Number of names/unique words/phoneme,Total speech time, Speech rate (number of words/total time), Spontaneous speech duration, Number of pauses, Speech tempo (number of phonemes per second),Number of nouns/verbs, Number of filler words/words, Pause time, Number of long pauses (≥2 seconds), Number of short pauses (<2 seconds), Number of filler pauses, Targeted speech/total speech). Ethics committee approval was obtained from the AUTF human research ethics committee (Decision No:İ05-280-22/ 12.05.2022).
RESULTS: A total of 105 participants, including 41 AH-MND, 29 MiND and 35 healthy elderly, were included in the study. When the groups were compared in terms of the scores of MMSE, MOCA, CDT, ÖAVLT, ACRT, Trail Making Test B-A, it was found that there was a significant difference between the three groups. The total recording time of 105 participants was 108.169 seconds. The speech parameters of the groups’ recordings are detailed in Table 1. The total pause duration of the AH-MND group was significantly longer than that of the MiND and healthy elderly groups. Although the number of words, unique words, nouns and phonemes used by the healthy elderly group during total speech is higher than the AH-MND and MiND groups, this difference is not statistically significant. AH-MND group used more filler words than MiND and healthy elderly group, but it was not statistically significant. There was a significant difference between AH-MND, MiND and control groups in the parameters of speech duration, storytelling duration, pause duration, number of long pauses, number of long pauses in spontaneous speech; there was a significant difference between AH-MND, MiND and healthy elderly groups in the parameters of speaking rate and paused speech rate in spontaneous speech; There was a significant difference between AD-MND and healthy groups in the tempo of speech during spontaneous speech (Table 2).
CONCLUSIONS: The findings obtained indicated that spontaneous speech characteristics were parallel to neuropsychological test performances, impairments in various components were observed in AH-MND and MiND cases, and some components were preserved. The main finding of this study was that the AH-MND group had significantly longer speech durations, but the amount of targeted speech and speech tempo were found to be lower. These findings support that speech-based methods can be developed as a cost-effective, non-invasive, and accessible diagnostic tool for the early diagnosis of Alzheimer’s disease. REFERENCES Ivanova, O., Martinez-Nicolas, I. & Meilan, J.J.G. (2024, Jan-Feb). Speech changes in old age: Methodological considerations for speech-based discrimination of healthy ageing and Alzheimer’s disease. Int J Lang Commun Disord, 59(1), 13-37. https://doi. org/10.1111/1460-6984.12888 Sanz, C., Carrillo, F., Slachevsky, A., Forno, G., Gorno Tempini, M.L., Villagra, R., Ibanez, A., Tagliazucchi, E. & Garcia, A.M. (2022). Automated text-level semantic markers of Alzheimer’s disease. Alzheimers Dement (Amst), 14(1), e12276. https://doi. org/10.1002/dad2.12276 Vigo, I., Coelho, L. & Reis, S. (2022, Jan 11). Speech- and Language- Based Classification of Alzheimer’s Disease: A Systematic Review. Bioengineering (Basel), 9(1). https://doi.org/10.3390/ bioengineering9010027 Keywords: Alzheimer’s disease, mild cognitive impairment, speech analysis