Open Access System for Information Sharing

Login Library

 

Thesis
Cited 0 time in webofscience Cited 0 time in scopus
Metadata Downloads

A Quantitative Assessment Methodology of Pharyngeal Swallow

Title
A Quantitative Assessment Methodology of Pharyngeal Swallow
Authors
이백희
Date Issued
2016
Publisher
포항공과대학교
Abstract
Dysphagia is the disorder related to difficulty in swallowing and is more frequently observed among patients with neurologic diseases such as stroke or people aged 65 and over. Undiagnosed or untreated dysphagia may lead to aspiration, pneumonia, dehydration, malnutrition, or even asphyxiation and death; therefore, early identification and appropriate treatment of dysphagia are important. Dysphagia has been mainly diagnosed by videofluoroscopic swallowing study (VFSS) and fiberoptic endoscopic evaluation of swallowing (FEES); but, both VFSS and FEES have limitations in terms of qualitative analysis and low safety (VFSS: radiation, FEES: invasion). The precedent study (Lee et al., 2012) developed a swallowing measurement device employing ultrasonic Doppler to measure a pharyngeal movement during swallowing. As a follow-up study, quantification, analysis, and interpretation of the measured pharyngeal swallowing signal and specialization for diagnosis of dysphagia are needed. The objective of the present study is to develop a quantitative assessment methodology of pharyngeal swallow. The specific objectives are as follows: (1) screening of swallowing out of pharyngeal movement signals, (2) establishment of a swallowing quantification protocol, (3) interpretation of pharyngeal movement signal by interoperating with VFSS video, (4) comparison of pharyngeal movement between healthy adults and dysphagic patients, and (5) development of a diagnostic model for discriminating the severity of dysphagia. First, a swallowing screening algorithm was developed to extract swallowing movements only out of various pharyngeal movements such as cough and vocalization measured by ultrasonic Doppler. Utilizing a concept of swallowing apnea that vocalization is impossible during pharyngeal swallow, a microphone for measuring audio signals was interoperated with the ultrasonic Doppler sensor. Signal processing techniques (e.g., moving average) and statistical methods (e.g., maximum-likelihood function) were also incorporated into the swallowing screening algorithm. As a result, the swallowing movement was completely discriminated from cough and vocalization that are also involved with pharyngeal movement. Second, to represent the characteristics of the pharyngeal movement during swallowing, the present study proposed five quantitative measures such as peak amplitude, duration time, number of peaks, peak-to-peak interval, and impulse by applying the four-step swallowing signal processing technique (S1. rectification, S2. smoothing, S3. peak detection, and S4. starting/ending points detection). A program that automatically calculates aforementioned five measures for a given swallowing signal was also developed. Third, pharyngeal movement signals were interpreted by interoperating with the corresponding VFSS video. The majority of pharyngeal movement signals showed two peaks, and 1st and 2nd peaks indicates ascending and descending movements of the laryngopharynx during swallowing, respectively. Based on VFSS video analysis, five measures of swallowing movement were interpreted as follows: peak amplitude – maximum instant movement of the laryngopharynx; duration time – total movement time in the laryngopharynx; number of peaks – number of movement changes in the laryngopharynx; peak-to-peak interval – bolus transportation time in the pharyngeal stage; and impulse – total movement of the laryngopharynx. Fourth, swallowing characteristics of healthy adults and dysphagic patients were analyzed by swallowing experiment. Swallowing signals for dry saliva, thin liquid 1, 3, 9 ml, thick liquid 1, 3, 9 ml were acquired from 120 healthy adults and 36 dysphagic patients. The swallowing signals from 88% of healthy adults showed one peak (49%) or two peaks (39%). Healthy adults were categorized as short-double peak (duration < 1 s and # peaks = 2; 43%), short-single peak (< 1 s and 1; 39%), and short-multiple peak (< 1 s and ≥ 3; 18%); dysphagic patients as short-double peak (< 1 s and 2; 58%), long-double peak (≥ 1 s and 2; 33%), and long-multiple peak (≥ 1 s and ≥ 3; 9%). Gender (F:M = 1:0.8), swallowing type (thick:thin = 1:1.2), and volume (1:3:9 ml = 1:1.1:1.3) were found significant on highest peak amplitude; swallowing type (thick:thin = 1:1.4) and volume (1:3:9 ml = 1:1.1:1.3) on impulse. Peak amplitude of dysphagic patients was 0.7 times lower compared with that of healthy adults; duration time 2.6 times longer; number of peaks 1.7 times higher; peak-to-peak interval 4.3 times longer; and impulse 0.8 times lower. Lastly, diagnostic models for discriminating the severity of dysphagia into normal, mild, and moderate/severe were developed. Five cumulative logit models for swallowing dry saliva, thin liquid 1 ml, 3 ml, thick liquid 1 ml, and 3 ml were developed using swallowing data of 120 healthy adults (normal) and 31 dysphagic patients (mild for 18 and moderate/severe for 13). The cumulative logit model for swallowing thin liquid 1 ml (input variables: age, gender, duration time, number of peaks, longest peak-to-peak interval, and impulse; sensitivity for mild = 50%, sensitivity for moderate/severe = 92%, specificity = 100%, and accuracy = 81%) was selected as the best model in terms of discriminant performances and practicality in clinics. The quantitative assessment methodology of the laryngopharyngeal movement during swallowing developed in the present study can contribute to real-time, accurate, and effective evaluation of the pharyngeal swallow. The swallowing analyses results of the laryngopharyngeal movement of healthy adults and dysphagic patients and the diagnostic model for discriminating the severity of dysphagia are readily applicable to medical diagnosis of dysphagia with VFSS in clinics.
삼킴 장애(연하 곤란, dysphagia)는 음식을 먹는 과정에서 발생하는 어려움의 질병으로서 주로 신경계통 질병 환자와 65세 이상 노년층에서 유병률이 높다. 삼킴 장애는 주로 흡인(aspiration), 폐렴(pneumonia), 탈수(dehydration), 영양실조(malnutrition) 등을 유발하고 심해지면 사망에 이를 수 있어 정확하고 신속한 진단이 중요하다. 기존 삼킴 장애는 주로 비디오 투시 조영 검사(videofluorocopic swallowing study, VFSS)와 비디오 내시경 검사(fiberoptic endoscopic evaluation of swallowing, FEES)를 통하여 진단되고 있으나 낮은 안전성(VFSS: 방사능, FEES: 침습)과 육안 평가의 한계가 있다. 선행 연구(Lee et al., 2012)에서는 삼킴 측정에 특화된 장비로서 인체에 무해한 ultrasonic Doppler를 사용하여 삼킴 시 인두(pharynx)의 움직임을 측정하는 장비가 개발되었는데, 측정된 인두 삼킴(pharyngeal swallow) 신호 중 삼킴 움직임만을 선별, 정량화, 분석, 해석하고 삼킴 장애 진단에 특화 시키는 연구가 필요하다. 본 연구는 ultrasonic Doppler를 사용하여 측정되는 인두 삼킴 신호를 사용하여 삼킴을 정량적으로 평가하기 위하여 (1) 인두 움직임 신호 중 삼킴 선별, (2) 삼킴 정량화 protocol 정립, (3) 삼킴 신호와 인두 움직임의 연동 해석, (4) 정상인과 삼킴 장애 환자의 삼킴 특성 비교 분석, 그리고 (5) 삼킴 장애 판별 모형 개발의 다섯 가지 세부 연구를 수행하였다. 첫째, ultrasonic Doppler로 측정되는 다양한 인두 움직임 중 삼킴 움직임만이 선별하는 알고리즘이 개발되었다. 인두 삼킴 시 소리가 발생할 수 없는 삼킴 무호흡(swallowing apnea) 개념을 적용하기 위하여 ultrasonic Doppler sensor에 소리 신호 획득을 위한 microphone이 연동되었다. 인두 삼킴 시 인두 움직임이 발생하지만 소리 발생이 불가능한 개념 구현에 특화된 신호 처리(예: moving average)와 통계 기법(예: maximum-likelihood function)을 적용하여 삼킴 선별 알고리즘이 개발되었다. 효용성 평가 결과, 삼킴 선별 알고리즘은 삼킴과 기침 및 발성 같이 소리가 함께 발생하는 인두 움직임을 100% 구별하였다. 둘째, 삼킴 신호 분석에 특화된 신호 처리 기법이 개발되고 삼킴 정량화 척도가 정립되었다. 네 단계 삼킴 신호 처리 기법(S1. rectification, S2. smoothing, S3. peak detection, S4. starting/ending points detection)을 적용하여 삼킴 움직임 특성을 나타내는 다섯 가지 삼킴 정량화 척도(peak amplitude, duration time, number of peaks, peak-to-peak interval, 그리고 impulse)가 개발되었다. 삼킴 신호만 입력되면 삼킴 정량화 척도가 자동으로 추출되는 S/W도 개발되었다. 셋째, 삼킴 시 삼킴 신호와 VFSS 영상을 동시에 획득하여 삼킴 신호와 인두 움직임이 연동 해석되었다. 정상 삼킴 신호의 peak 개수는 대부분 2가지로 나타나며, 첫 번째 peak는 삼킴 시 인후두(laryngopharynx) 상승 움직임을 두 번째 peak는 삼킴 시 laryngopharynx 하강 움직임으로 파악되었다. VFSS 영상 분석 결과에 근간하여 peak amplitude는 laryngopharynx의 순간 최대 움직임 정도, duration time은 laryngopharynx의 총 움직임 시간, number of peaks는 laryngopharynx의 움직임 변환 횟수, peak-to-peak interval는 인두 삼킴 시 bolus 이동 시간, 그리고 impulse는 laryngopharynx의 총 움직임 정도로 의미가 정립되었다. 넷째, 정상인과 삼킴 장애 환자를 대상으로 삼킴 실험을 수행하여 삼킴 특성이 비교 분석되었다. 정상인 120명과 삼킴 장애 환자 36명에 대해 침, thin liquid 1, 3, 9 ml, thick liquid 1, 3, 9 ml 삼킴에 대한 삼킴 신호가 획득되었다. 정상인의 number of peaks는 1개(49%) 또는 2개(39%)일 때가 88%로 나타났다. 삼킴 정량화 척도에 대해 K-mean clustering을 적용(K = 3)한 별과, 정상인은 short-double peak (duration < 1 s and # peaks = 2)가 43%, short-single peak (duration < 1 s and # peaks = 1)가 39%, short-multiple peak (duration < 1 s and # peaks ≥ 3)가 18%로 나타났으며, 삼킴 장애 환자는 short-double peak (duration < 1 s and # peaks = 2)가 58%, long-double peak (duration ≥ 1 s and # peaks = 2)가 33%, long-multiple peak (duration ≥ 1 s and # peaks ≥ 3)가 9%인 것으로 나타났다. 정상인의 highest peak amplitude에는 성별(female:male = 1:0.8), 삼킴 종류(thick liquid:thin liquid = 1:1.2), 삼킴 용량(1 ml:3 ml:9 ml = 1:1.1:1.3)이 유의하였으며, impulse에는 삼킴 종류(thick liquid:thin liquid = 1:1.4)와 삼킴 용량(1 ml:3 ml:9 ml = 1:1.1:1.3)이 유의한 것으로 나타났다. 삼킴 장애 환자는 정상인에 비해 평균적으로 peak amplitude가 0.7배 낮고, duration time이 2.6배 길고, number of peaks가 1.7배 많고, peak-to-peak interval이 4.3배 길고, 그리고 impulse가 0.8배 낮은 것으로 나타났다. 마지막으로, 삼킴 장애 심각도를 normal, mild, 그리고 moderate/severe로 분류하는 삼킴 장애 판별 모형이 개발되었다. 정상인 120명과 삼킴 장애 환자 31명(VFSS 진단 결과: mild 18명, moderate/severe: 13명)의 침, thin liquid 1 ml, 3 ml, thick liquid 1 ml, 그리고 3 ml 삼킴에 대한 5가지 cumulative logit model이 개발되었다. 최적 모형은 판별 성능과 실용성이 우수한 thin liquid 1 ml 삼킴에 대한 cumulative logit model (입력 변수: 연령, 성별, duration time, number of peaks, longest peak-to-peak interval, impulse; 판별 성능: sensitivity for mild = 50%, sensitivity for moderate/severe = 92%, specificity = 100%, accuracy = 81%)로 선정되었다. 본 연구의 ultrasonic Doppler를 사용한 삼킴 시 laryngopharynx 움직임 정량화 방법론은 인두 삼킴을 실시간 정확하고 효과적으로 평가하는데 기여할 수 있다. 본 연구의 정상인과 삼킴 장애 환자의 삼킴 시 laryngopharynx 움직임 특성 비교 분석 결과와 삼킴 장애 심각도 판별 모형은 임상에서의 VFSS 검사와 더불어 삼킴 장애의 과학적 진단에 적용될 수 있을 것으로 기대된다.
URI
http://postech.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000002224314
https://oasis.postech.ac.kr/handle/2014.oak/92810
Article Type
Thesis
Files in This Item:
There are no files associated with this item.

qr_code

  • mendeley

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Views & Downloads

Browse