Date on Master's Thesis/Doctoral Dissertation

5-2022

Document Type

Master's Thesis

Degree Name

M. Eng.

Department

Bioengineering

Degree Program

JB Speed School of Engineering

Committee Chair

Kopechek, Jonathan

Committee Co-Chair (if applicable)

Frieboes, Hermann

Committee Member

Frieboes, Hermann

Committee Member

Yaddanapudi, Kavitha

Author's Keywords

Machine Learning; Acoustic Attenuation; Cellular Characterization

Abstract

T-cell therapies have been gaining popularity in recent years due to their cancer fighting potential. With remission rates improving in this field of immunotherapy, the demand for T-cell therapies has also increased; however, the cell processing techniques for these therapeutic products have yet to rise to the level of demand. The manufacturing process takes too long and a significant amount of processed cell batches can fail to meet safety requirements. These limitations of cell processing can be detrimental to patients seeking T-cell therapies. While current products have improved the time it takes to manufacture these therapeutic products, there is still a lack of an in-line non-destructive sample quality control monitoring system to reduce the risk of batch failures and delays. In this thesis, theoretical and experimental testing was conducted to serve as a proof of concept that machine learning analysis of acoustic attenuation signals could be utilized for cellular characterization. A machine learning analysis method was able to determine sizes of theoretical microparticles and concentrations of different cell lines from acoustic attenuation signals collected in a static ultrasound chamber, as well as a continuous flow ultrasound chamber. It was found that the machine learning technique called scratch learning generally served as a better model for these cellular characterization trials, rather than transfer learning. With further refinement of the scratch learning architecture, as well as the further development of the attenuation signal acquisition system, optimization of classification accuracy of the machine learning analysis method could be further improved. This optimization could enable an in-line ultrasound-based quality control module for classification of multiple cell characteristics to be implemented into cell processing procedures for T-cell therapies.

Download

Share

COinS