Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon
Physiological signals play a significant role in clinical diagnosis, it always acts as a major input of a decision support system. However, the physiological signal is easily corrupted by different factors especially motion artifacts. Several research works have been tried to recover the underlying...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/12501/ http://studentsrepo.um.edu.my/12501/1/Lim_Pooi_Khoon.pdf http://studentsrepo.um.edu.my/12501/2/Lim_Pooi_Khoon.pdf |
| _version_ | 1848774670684258304 |
|---|---|
| author | Lim , Pooi Khoon |
| author_facet | Lim , Pooi Khoon |
| author_sort | Lim , Pooi Khoon |
| building | UM Research Repository |
| collection | Online Access |
| description | Physiological signals play a significant role in clinical diagnosis, it always acts as a major input of a decision support system. However, the physiological signal is easily corrupted by different factors especially motion artifacts. Several research works have been tried to recover the underlying physiological signal by suppressing the artifact. However, not much attention has been paid to situation where the artifact is too extreme and the artifact suppression is not possible. In this situation, physiological signal quality must be evaluated before any further assessment. In this study, an automated artifact detection algorithm was developed for Blood Pressure and PPG signals. For Blood Pressure signal, an automatic algorithm based on relative changes in the cuff pressure and neighbouring oscillometric pulses was proposed to remove outlier points caused by movement artifacts. Next, multiple linear regression (MLR) and support vector regression (SVR) models were used to examine the relationship between the Systolic Blood Pressure (SBP) and the Diastolic Blood Pressure (DBP) ratio with ten features extracted from the oscillometric waveform envelope (OWE). Upon using the artifact detection method followed by BP estimation, the SBP and DBP were improved in BHS grades from D to A. With regards to the AAMI standard, the mean ± SD of difference between the estimated and the gold standard SBP improved from 4.5±28.6 mmHg to -0.3±5.8mmHg and -0.6±5.4 mmHg using the MLR and SVR, respectively. Meanwhile, the mean ± SD of difference for DBP improved from 0.0±14.2 mmHg to -0.2±6.4 mmHg and 0.4±6.3 mmHg using the MLR and SVR, respectively. For PPG signal, two master templates have been generated from PhysioNet MIMIC II database. The master template is then updated with each of the incoming clean pulse. Correlation coefficient were used to classify the PPG pulse into either good or bad quality categories. The robustness of this artifact detection algorithm was then evaluated on both short and continuous data collected from young and older subjects which included arrhythmia patients. For short data, the average accuracy improved from 95.2% to 98.0%. For long continuous data on healthy subject, an average accuracy of 91.5%, sensitivity of 94.1% and specificity of 89.7% were achieved. Meanwhile, for long continuous data on elder subject which included arrhythmia patients, an average accuracy of 91.3%, sensitivity of 80.5% and specificity 93.0% were achieved.
|
| first_indexed | 2025-11-14T14:02:00Z |
| format | Thesis |
| id | um-12501 |
| institution | University Malaya |
| institution_category | Local University |
| last_indexed | 2025-11-14T14:02:00Z |
| publishDate | 2020 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | um-125012023-01-17T23:50:00Z Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon Lim , Pooi Khoon T Technology (General) TA Engineering (General). Civil engineering (General) Physiological signals play a significant role in clinical diagnosis, it always acts as a major input of a decision support system. However, the physiological signal is easily corrupted by different factors especially motion artifacts. Several research works have been tried to recover the underlying physiological signal by suppressing the artifact. However, not much attention has been paid to situation where the artifact is too extreme and the artifact suppression is not possible. In this situation, physiological signal quality must be evaluated before any further assessment. In this study, an automated artifact detection algorithm was developed for Blood Pressure and PPG signals. For Blood Pressure signal, an automatic algorithm based on relative changes in the cuff pressure and neighbouring oscillometric pulses was proposed to remove outlier points caused by movement artifacts. Next, multiple linear regression (MLR) and support vector regression (SVR) models were used to examine the relationship between the Systolic Blood Pressure (SBP) and the Diastolic Blood Pressure (DBP) ratio with ten features extracted from the oscillometric waveform envelope (OWE). Upon using the artifact detection method followed by BP estimation, the SBP and DBP were improved in BHS grades from D to A. With regards to the AAMI standard, the mean ± SD of difference between the estimated and the gold standard SBP improved from 4.5±28.6 mmHg to -0.3±5.8mmHg and -0.6±5.4 mmHg using the MLR and SVR, respectively. Meanwhile, the mean ± SD of difference for DBP improved from 0.0±14.2 mmHg to -0.2±6.4 mmHg and 0.4±6.3 mmHg using the MLR and SVR, respectively. For PPG signal, two master templates have been generated from PhysioNet MIMIC II database. The master template is then updated with each of the incoming clean pulse. Correlation coefficient were used to classify the PPG pulse into either good or bad quality categories. The robustness of this artifact detection algorithm was then evaluated on both short and continuous data collected from young and older subjects which included arrhythmia patients. For short data, the average accuracy improved from 95.2% to 98.0%. For long continuous data on healthy subject, an average accuracy of 91.5%, sensitivity of 94.1% and specificity of 89.7% were achieved. Meanwhile, for long continuous data on elder subject which included arrhythmia patients, an average accuracy of 91.3%, sensitivity of 80.5% and specificity 93.0% were achieved. 2020-05 Thesis NonPeerReviewed application/pdf http://studentsrepo.um.edu.my/12501/1/Lim_Pooi_Khoon.pdf application/pdf http://studentsrepo.um.edu.my/12501/2/Lim_Pooi_Khoon.pdf Lim , Pooi Khoon (2020) Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon. PhD thesis, Universiti Malaya. http://studentsrepo.um.edu.my/12501/ |
| spellingShingle | T Technology (General) TA Engineering (General). Civil engineering (General) Lim , Pooi Khoon Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title | Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title_full | Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title_fullStr | Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title_full_unstemmed | Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title_short | Artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / Lim Pooi Khoon |
| title_sort | artifact identification for blood pressure and photoplethysmography signals in an unsupervised environment / lim pooi khoon |
| topic | T Technology (General) TA Engineering (General). Civil engineering (General) |
| url | http://studentsrepo.um.edu.my/12501/ http://studentsrepo.um.edu.my/12501/1/Lim_Pooi_Khoon.pdf http://studentsrepo.um.edu.my/12501/2/Lim_Pooi_Khoon.pdf |