서브메뉴
검색
Automated ELISA and Machine Learning-Enhanced Accuracy in Biosensing and Colorimetric Measurements
Automated ELISA and Machine Learning-Enhanced Accuracy in Biosensing and Colorimetric Measurements
Detailed Information
- Material Type
- 단행본
- 0017359879
- Date and Time of Latest Transaction
- 20260202105230
- ISBN
- 9798291567180
- DDC
- 610
- Author
- Aalizadeh, Majid.
- Title/Author
- Automated ELISA and Machine Learning-Enhanced Accuracy in Biosensing and Colorimetric Measurements
- Publish Info
- [Sl] : University of Michigan, 2025
- Publish Info
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- Material Info
- 186 p
- General Note
- Source: Dissertations Abstracts International, Volume: 87-02, Section: B.
- General Note
- Advisor: Fan, Xudong;Guo, L. Jay.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- Abstracts/Etc
- 요약Modern biomedical diagnostics demand platforms that are accurate, compact, automated, and cost-effective. Conventional ELISA (enzyme-linked immunosorbent assay) systems, while reliable, are often bulky, manual, and expensive, limiting accessibility for decentralized testing. This dissertation addresses this gap by developing a miniaturized ELISA platform with a production cost of about $1200, integrating simple rotary and linear movements as the only required mechanical motions with custom three-dimensional printed components to automate liquid handling. Using interleukin-6 (IL-6) as a model system, the platform achieved an R² of 0.9937 with duplicate measurements and a limit of detection (LOD) of 7.13 pg/mL, matching or exceeding the performance of high-end commercial systems while offering significant reductions in size, complexity, and cost. Beyond automation, this dissertation focuses on leveraging machine learning to enhance optical biosensing accuracy without modifying the hardware. In both primary sensing methods, which are peak shift tracking and fixed-wavelength intensity modulation, traditional approaches rely on one-dimensional fitting and overlook the rich spectral information available. First, a Ridge Regression framework is applied to combine multiple resonant peak shifts in silicon nanorod arrays, achieving approximately three orders of magnitude improvement in mean squared error (MSE) of refractive index prediction compared to traditional single-peak tracking. Full-spectrum modeling is then explored across two optical structures, one based on silicon supporting sharp Mie resonances for peak shift-based sensing and another based on titanium where intensity modulation is the dominant response to the refractive index change. It is found that linear regression based full-spectrum analysis significantly enhances accuracy for intensity modulation-based index sensing, while for peak shift-based sensing, combining multiple peaks provides superior performance. Finally, experimental validation using colorimetric measurements shows that applying machine learning with twelve selected wavelengths results in more than a 5700-fold MSE reduction relative to the best single-wavelength result.
- Subject Added Entry-Topical Term
- Biomedical engineering
- Subject Added Entry-Topical Term
- Engineering
- Subject Added Entry-Topical Term
- Nanotechnology
- Index Term-Uncontrolled
- Machine learning
- Index Term-Uncontrolled
- Biomedical diagnostics
- Index Term-Uncontrolled
- Biosensing
- Index Term-Uncontrolled
- Limit of detection
- Index Term-Uncontrolled
- Colorimetric measurements
- Added Entry-Corporate Name
- University of Michigan Electrical and Computer Engineering
- Host Item Entry
- Dissertations Abstracts International. 87-02B.
- Electronic Location and Access
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359879
■00520260202105230
■006m o d
■007cr#unu||||||||
■020 ▼a9798291567180
■035 ▼a(MiAaPQ)AAI32271883
■035 ▼a(MiAaPQ)umichrackham006389
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a610
■1001 ▼aAalizadeh, Majid.
■24510▼aAutomated ELISA and Machine Learning-Enhanced Accuracy in Biosensing and Colorimetric Measurements
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a186 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-02, Section: B.
■500 ▼aAdvisor: Fan, Xudong;Guo, L. Jay.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aModern biomedical diagnostics demand platforms that are accurate, compact, automated, and cost-effective. Conventional ELISA (enzyme-linked immunosorbent assay) systems, while reliable, are often bulky, manual, and expensive, limiting accessibility for decentralized testing. This dissertation addresses this gap by developing a miniaturized ELISA platform with a production cost of about $1200, integrating simple rotary and linear movements as the only required mechanical motions with custom three-dimensional printed components to automate liquid handling. Using interleukin-6 (IL-6) as a model system, the platform achieved an R² of 0.9937 with duplicate measurements and a limit of detection (LOD) of 7.13 pg/mL, matching or exceeding the performance of high-end commercial systems while offering significant reductions in size, complexity, and cost. Beyond automation, this dissertation focuses on leveraging machine learning to enhance optical biosensing accuracy without modifying the hardware. In both primary sensing methods, which are peak shift tracking and fixed-wavelength intensity modulation, traditional approaches rely on one-dimensional fitting and overlook the rich spectral information available. First, a Ridge Regression framework is applied to combine multiple resonant peak shifts in silicon nanorod arrays, achieving approximately three orders of magnitude improvement in mean squared error (MSE) of refractive index prediction compared to traditional single-peak tracking. Full-spectrum modeling is then explored across two optical structures, one based on silicon supporting sharp Mie resonances for peak shift-based sensing and another based on titanium where intensity modulation is the dominant response to the refractive index change. It is found that linear regression based full-spectrum analysis significantly enhances accuracy for intensity modulation-based index sensing, while for peak shift-based sensing, combining multiple peaks provides superior performance. Finally, experimental validation using colorimetric measurements shows that applying machine learning with twelve selected wavelengths results in more than a 5700-fold MSE reduction relative to the best single-wavelength result.
■590 ▼aSchool code: 0127.
■650 4▼aBiomedical engineering
■650 4▼aEngineering
■650 4▼aNanotechnology
■653 ▼aMachine learning
■653 ▼aBiomedical diagnostics
■653 ▼aBiosensing
■653 ▼aLimit of detection
■653 ▼aColorimetric measurements
■690 ▼a0541
■690 ▼a0537
■690 ▼a0652
■690 ▼a0800
■71020▼aUniversity of Michigan▼bElectrical and Computer Engineering.
■7730 ▼tDissertations Abstracts International▼g87-02B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359879▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
Preview
Export
ChatGPT Discussion
AI Recommended Related Books
Detail Info.
- Reservation
- Not Exist
- My Folder
- First Request
- Non-Book Loan Application
- Nighttime Book Loan Application
Available after logging in.


