Multiple sclerosis (Houndmills, Basingstoke, England) · 2024
From "invisible" to "audible": Features extracted during simple speech tasks classify patient-reported fatigue in multiple sclerosis.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Nylander Alyssa, Sisodia Nikki, Henderson Kyra, Wijangco Jaeleene, Koshal Kanishka, Poole Shane, Dias Marcelo, Linz Nicklas, Tröger Johannes, König Alexandra, Hayward-Koennecke Helen, Pedotti Rosetta, Brown Ethan, Halabi Cathra, Staffaroni Adam, Bove Riley
Аннотация
BACKGROUND: Fatigue is a major "invisible" symptom in people with multiple sclerosis (PwMS), which may affect speech. Automated speech analysis is an objective, rapid tool to capture digital speech biomarkers linked to functional outcomes.
OBJECTIVE: To use automated speech analysis to assess multiple sclerosis (MS) fatigue metrics.
METHODS: Eighty-four PwMS completed scripted and spontaneous speech tasks; fatigue was assessed with Modified Fatigue Impact Scale (MFIS). Speech was processed using an automated speech analysis pipeline (ki elements: SIGMA speech processing library) to transcribe speech and extract features. Regression models assessed associations between speech features and fatigue and validated in a separate set of 30 participants.
RESULTS: Cohort characteristics were as follows: mean age 49.8 (standard deviation (SD) = 13.6), 71.4% female, 85% relapsing-onset, median Expanded Disability Status Scale (EDSS) 2.5 (range: 0-6.5), mean MFIS 27.6 (SD = 19.4), and 30% with MFIS > 38. MFIS moderately correlated with pitch (R = 0.32, p = 0.005), pause duration (R = 0.33, p = 0.007), and utterance duration (R = 0.31, p = 0.0111). A logistic model using speech features from multiple tasks accurately classified MFIS in training (area under the curve (AUC) = 0.95, R2 = 0.59, p < 0.001) and test sets (AUC = 0.93, R2 = 0.54, p = 0.0222). Adjusting for EDSS, processing speed, and depression in sensitivity analyses did not impact model accuracy.
CONCLUSION: Fatigue may be assessed using simple, low-burden speech tasks that correlate with gold-standard subjective fatigue measures.
OBJECTIVE: To use automated speech analysis to assess multiple sclerosis (MS) fatigue metrics.
METHODS: Eighty-four PwMS completed scripted and spontaneous speech tasks; fatigue was assessed with Modified Fatigue Impact Scale (MFIS). Speech was processed using an automated speech analysis pipeline (ki elements: SIGMA speech processing library) to transcribe speech and extract features. Regression models assessed associations between speech features and fatigue and validated in a separate set of 30 participants.
RESULTS: Cohort characteristics were as follows: mean age 49.8 (standard deviation (SD) = 13.6), 71.4% female, 85% relapsing-onset, median Expanded Disability Status Scale (EDSS) 2.5 (range: 0-6.5), mean MFIS 27.6 (SD = 19.4), and 30% with MFIS > 38. MFIS moderately correlated with pitch (R = 0.32, p = 0.005), pause duration (R = 0.33, p = 0.007), and utterance duration (R = 0.31, p = 0.0111). A logistic model using speech features from multiple tasks accurately classified MFIS in training (area under the curve (AUC) = 0.95, R2 = 0.59, p < 0.001) and test sets (AUC = 0.93, R2 = 0.54, p = 0.0222). Adjusting for EDSS, processing speed, and depression in sensitivity analyses did not impact model accuracy.
CONCLUSION: Fatigue may be assessed using simple, low-burden speech tasks that correlate with gold-standard subjective fatigue measures.