Bian, Long
(2021)
APPLICATION OF MACHINE LEARNING IN CARBON NANOTUBE-BASED BIOSENSING.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Carbon nanotubes have attracted a lot of attention since their electron microscopy images were published 30 years ago. Among many applications, carbon nanotubes are uniquely suited for development of biosensors due to high aspect ratio and high charge mobility of the carbon nanotubes. Carbon nanotube-based field-effect transistors (NTFETs) have been extensively studied as chemical sensors in medical, safety, environmental, and provide rich information regarding carbon nanotube interactions with target analytes. Being able to leverage the wealth of information extracted from NTFET characteristics will make it possible to 1) better select the sensor surface functionalization for sensing target, 2) investigate the interactions between sensor surface and analytes, 3) elucidate the sensing mechanism. Machine learning is a very powerful tool as it specializes in extracting useful information from dataset, especially high dimensional data produced by NTFETs, which can be further utilized for constructing classification models to predict the presence of biomarker and regression models to quantify the biomarker concentrations in the environment.
In this dissertation, both classification models and regression models were constructed to meet different sensing tasks. In classification models, both linear discriminant analysis (LDA) and support vector classifier with linear kernel (SVC) models were constructed to predict the presence of caffeine in different solutions and commercial beverages, cross-validated accuracy was utilized as a metric to evaluate predictability of the models. Recursive feature elimination and correlation analysis were performed to evaluate the feature importance and selected the best combination of v features to give the best performing models. In regression, a series of regression models were constructed to capture the non-linearity in the data as well as accommodate the model complexity, and the sensors dynamic range was greatly extended. By analyzing the feature importance in all the models and performing control experiment, the proposed sensing mechanism after the receptor’s saturation was validated. Finally, an automated lab testing system was designed, constructed, and fully implemented, to accelerate the sensing testing and data collection workflow. We also utilized the automated system to perform high-throughput nanomaterial screening for NTFET-based sensors.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
18 December 2021 |
| Date Type: |
Publication |
| Defense Date: |
15 November 2021 |
| Approval Date: |
13 September 2024 |
| Submission Date: |
5 December 2021 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
165 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Dietrich School of Arts and Sciences > Chemistry |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Biosensor, Machine Learning, Carbon Nanomaterials |
| Date Deposited: |
13 Sep 2024 18:57 |
| Last Modified: |
13 Sep 2024 19:07 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/42006 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}