Machine Learning-Based Quantification of Lateral Flow Assay Using Smartphone-Captured Images.

Area of Science:

• Biomedical diagnostics
• Machine learning applications
• Image analysis in healthcare

Background:

• Lateral flow assays (LFAs) are widely used for convenient, low-cost point-of-care diagnostics.
• Traditional LFAs lack precise quantification, offering only qualitative (yes/no) results.
• Improving LFA quantification is crucial for accurate disease detection and monitoring.

Purpose of the Study:

• To develop and evaluate machine learning (ML) and deep learning (DL) models for quantifying analyte levels in LFAs.
• To analyze smartphone-captured LFA images for improved diagnostic accuracy.
• To compare the performance of different ML/DL models in LFA result interpretation.

Main Methods:

• Development of random forest and convolutional neural network (CNN) models.
• Training and testing models on smartphone-captured images of LFAs.
• Comparative analysis of model performance based on image size and noise levels.

Main Results:

• Both random forest and CNN models demonstrated strong performance in classifying LFA results.
• Random forest models outperformed CNNs when trained on small-size images.
• CNN models showed superior performance in classifying noisy LFA images.

Conclusions:

• ML and DL models can effectively quantify analyte loads from LFA images, enhancing diagnostic capabilities.
• The choice of model (random forest vs. CNN) depends on image characteristics (size, noise).
• This approach offers a pathway to more accurate and quantitative point-of-care diagnostics using smartphones.