Classification of nucleic acid amplification on ISFET arrays using spectrogram-based neural networks.

Area of Science:

• Biotechnology
• Artificial Intelligence
• Molecular Diagnostics

Background:

• The COVID-19 pandemic revealed critical needs in molecular diagnostics for rapid, private, and secure solutions.
• Existing methods often lack the speed and portability required for point-of-care applications.
• There is a growing demand for AI-driven edge computing solutions in healthcare.

Purpose of the Study:

• To present a proof-of-concept for detecting nucleic acid amplification using ISFET sensors and deep learning.
• To develop a low-cost, portable lab-on-chip platform for identifying infectious diseases and cancer biomarkers.
• To enhance molecular diagnostic capabilities through AI-based edge solutions.

Main Methods:

• Utilized Ion-Sensitive Field-Effect Transistor (ISFET) sensors for nucleic acid detection.
• Applied deep learning, specifically 2D convolutional neural networks, for signal classification.
• Transformed sensor signals into spectrograms (time-frequency domain) for improved AI compatibility and performance.

Main Results:

• Achieved reliable classification of chemical signals using image processing techniques on spectrograms.
• Demonstrated significant performance improvement compared to time-domain data analysis.
• Developed a deep learning model with 84% accuracy and a compact 30kB size, suitable for edge devices.

Conclusions:

• The developed method enables rapid and reliable molecular diagnostics on a portable platform.
• Spectrogram transformation enhances AI model performance for chemical signal analysis.
• This approach paves the way for intelligent, AI-enabled lab-on-chip systems for advanced diagnostics.