Machine learning-empowered smartphone platforms for sustainable point-of-care bacterial detection.

Area of Science:

• Biomedical Engineering
• Computer Science
• Analytical Chemistry

Background:

• Smartphone biosensing offers decentralized bacterial detection via miniaturized optics and electronics.
• Sensing signals are often high-dimensional and noisy, affected by environmental factors.
• Machine learning (ML) can address these challenges by extracting features and learning patterns.

Purpose of the Study:

• To review recent developments in ML-enabled smartphone biosensing for bacterial identification.
• To cover various ML techniques, platforms, and data workflows.
• To discuss translational barriers and future directions for reliable diagnostics.

Main Methods:

• Survey of supervised learning (SVM, KNN, random forests) and deep learning (CNNs, YOLO, transformers, autoencoders).
• Integration of feature engineering and dimensionality reduction.
• Analysis of fluorescence, colorimetric, bioluminescent, electrochemical, and SERS-based systems.

Main Results:

• ML improves classification and quantification performance in bacterial sensing.
• Reviewed systems demonstrate enhanced accuracy, limit of detection, and robustness.
• Validation in real samples highlights the potential of these integrated approaches.

Conclusions:

• ML is crucial for overcoming limitations in smartphone-based bacterial detection.
• Addressing generalizability, interpretability, and data scarcity is key for clinical translation.
• Future work should focus on reliable, scalable, and clinically meaningful diagnostics.