國立虎尾科技大學 |

Electronic Nose Design for Respiratory Disease Detection : = Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Electronic Nose Design for Respiratory Disease Detection :/
Reminder of title:	Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm.
Author:	Guo, Xufeng.
Description:	1 online resource (131 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
Contained By:	Dissertations Abstracts International85-06B.
Subject:	Mechanical engineering. -
Online resource:	click for full text (PQDT)
ISBN:	9798380998086

Electronic Nose Design for Respiratory Disease Detection : = Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm.
Guo, Xufeng.

Electronic Nose Design for Respiratory Disease Detection :Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm. - 1 online resource (131 pages)

Source: Dissertations Abstracts International, Volume: 85-06, Section: B.

Thesis (Ph.D.)--The University of Akron, 2023.

Includes bibliographical references

E-nose-based exhaled breath analysis is becoming a promising method for non-invasive disease diagnostics, yet its wide commercial deployment demands advances in gas sensor technology, e-nose system design, as well as recognition algorithms. This work is based on the workflow of E-nose and focuses on the design and optimization of the E-nose system for three sub-modules, including the flow delivery system, testing system, and recognition algorithms.A study of velocity-dependent adsorption system is investigated by both experiment and model. In the test of NH3 detection, the flow velocity shows a different effect on sensor response with varying concentrations. For low concentrations, an increase in velocity enhances the sensor's sensitivity. However, for high concentrations, the sensitivity initially increases and then decreases. The mechanism behind this phenomenon is examined by considering molecular diffusion and adsorption processes. To provide a deeper understanding and explanation, a model is developed based on the hypothesis that flow velocity promotes the desorption process and influences the concentration surrounding the sensor.A cutting-edge mobile electronic nose (e-nose) device, outfitted with a sophisticated sensor array and an optimized system for flow delivery, alongside online data transmission and machine learning algorithms, has been developed to proficiently identify and analyze volatile compounds in exhaled breath, thereby offering a systematic design strategy for non-invasive disease diagnosis. Central to this device, the gas sensor array is constructed with eight graphene-based chemo-resistive materials, each highly sensitive to trace gas analytes. In a digital breath test study, the device was able to distinguish patients with respiratory diseases from healthy individuals with an overall accuracy exceeding 80%. Such results suggest that the proposed portable e-nose prototype presents significant potential for personalized health monitoring.Inspired by the biological olfactory neural system, an exhaled breath gas analysis algorithm is proposed to classify respiratory diseases based on wavelet transform pre-processing leveraging transfer learning. Wavelet transform expands time series sensor data into two-dimensional temporal and frequency information, mimicking the function of olfactory receptor neurons (ORNs) to produce high-dimensional representations of chemical, spatial, and temporal information for downstream neural networks. The two- dimensional data is then fed into a pre-trained VGG16-based convolutional neural network (CNN) for disease recognition and classification. Reusing a pre-trained model, also called transfer learning, enables the capacity of deep learning with limited patient cases. Tested on a dataset with 383 respiratory disease patients. The biologically inspired algorithm achieves an accuracy of 85%, outperforming conventional machine learning algorithms, including support vector machines, random forests, gradient boosting trees, multiple layer perceptions (MLP), 1D-CNN, and Recurrent Neural Networks (RNN).The study of the velocity-dependent sensing mechanism, the E-nose device design and optimization, and the test and the biologically inspired network architecture imply that this research could be a reliable and scalable strategy for disease detection.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798380998086Subjects--Topical Terms:

557493
Mechanical engineering.
Subjects--Index Terms:

E-noseIndex Terms--Genre/Form:

554714
Electronic books.

Electronic Nose Design for Respiratory Disease Detection : = Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm.
LDR:04812ntm a2200397K 4500 001 1141672
005 20240318062712.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798380998086
035 $a (MiAaPQ)AAI30980431
035 $a (MiAaPQ)OhioLINKakron1698687743930654
035 $a AAI30980431
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Guo, Xufeng. $3 1465593
245 1 0 $a Electronic Nose Design for Respiratory Disease Detection : $b Flow Delivery, Sensing Mechanism, and Machine Learning Recognition Algorithm.
264 0 $c 2023
300 $a 1 online resource (131 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
500 $a Advisor: Dong, Yalin.
502 $a Thesis (Ph.D.)--The University of Akron, 2023.
504 $a Includes bibliographical references
520 $a E-nose-based exhaled breath analysis is becoming a promising method for non-invasive disease diagnostics, yet its wide commercial deployment demands advances in gas sensor technology, e-nose system design, as well as recognition algorithms. This work is based on the workflow of E-nose and focuses on the design and optimization of the E-nose system for three sub-modules, including the flow delivery system, testing system, and recognition algorithms.A study of velocity-dependent adsorption system is investigated by both experiment and model. In the test of NH3 detection, the flow velocity shows a different effect on sensor response with varying concentrations. For low concentrations, an increase in velocity enhances the sensor's sensitivity. However, for high concentrations, the sensitivity initially increases and then decreases. The mechanism behind this phenomenon is examined by considering molecular diffusion and adsorption processes. To provide a deeper understanding and explanation, a model is developed based on the hypothesis that flow velocity promotes the desorption process and influences the concentration surrounding the sensor.A cutting-edge mobile electronic nose (e-nose) device, outfitted with a sophisticated sensor array and an optimized system for flow delivery, alongside online data transmission and machine learning algorithms, has been developed to proficiently identify and analyze volatile compounds in exhaled breath, thereby offering a systematic design strategy for non-invasive disease diagnosis. Central to this device, the gas sensor array is constructed with eight graphene-based chemo-resistive materials, each highly sensitive to trace gas analytes. In a digital breath test study, the device was able to distinguish patients with respiratory diseases from healthy individuals with an overall accuracy exceeding 80%. Such results suggest that the proposed portable e-nose prototype presents significant potential for personalized health monitoring.Inspired by the biological olfactory neural system, an exhaled breath gas analysis algorithm is proposed to classify respiratory diseases based on wavelet transform pre-processing leveraging transfer learning. Wavelet transform expands time series sensor data into two-dimensional temporal and frequency information, mimicking the function of olfactory receptor neurons (ORNs) to produce high-dimensional representations of chemical, spatial, and temporal information for downstream neural networks. The two- dimensional data is then fed into a pre-trained VGG16-based convolutional neural network (CNN) for disease recognition and classification. Reusing a pre-trained model, also called transfer learning, enables the capacity of deep learning with limited patient cases. Tested on a dataset with 383 respiratory disease patients. The biologically inspired algorithm achieves an accuracy of 85%, outperforming conventional machine learning algorithms, including support vector machines, random forests, gradient boosting trees, multiple layer perceptions (MLP), 1D-CNN, and Recurrent Neural Networks (RNN).The study of the velocity-dependent sensing mechanism, the E-nose device design and optimization, and the test and the biologically inspired network architecture imply that this research could be a reliable and scalable strategy for disease detection.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Mechanical engineering. $3 557493
650 4 $a Biomedical engineering. $3 588770
653 $a E-nose
653 $a Artificial olfactory
653 $a Machine learning
653 $a Deep learning
653 $a Disease detection
655 7 $a Electronic books. $2 local $3 554714
690 $a 0548
690 $a 0541
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a The University of Akron. $b Mechanical Engineering. $3 1180233
773 0 $t Dissertations Abstracts International $g 85-06B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30980431 $z click for full text (PQDT)