A deep learning-driven low-power, accurate, and portable platform for rapid detection of COVID-19 using reverse-transcription loop-mediated isothermal amplification

Abstract

This paper presents a deep learning-driven portable, accurate, low-cost, and easy-to-use device to perform Reverse-Transcription Loop-Mediated Isothermal Amplification (RT-LAMP) to facilitate rapid detection of COVID-19. The 3D-printed device-powered using only a 5 Volt AC-DC adapter-can perform 16 simultaneous RT-LAMP reactions and can be used multiple times. Moreover, the experimental protocol is devised to obviate the need for separate, expensive equipment for RNA extraction in addition to eliminating sample evaporation. The entire process from sample preparation to the qualitative assessment of the LAMP amplification takes only 45 min (10 min for pre-heating and 35 min for RT-LAMP reactions). The completion of the amplification reaction yields a fuchsia color for the negative samples and either a yellow or orange color for the positive samples, based on a pH indicator dye. The device is coupled with a novel deep learning system that automatically analyzes the amplification results and pays attention to the pH indicator dye to screen the COVID-19 subjects. The proposed device has been rigorously tested on 250 RT-LAMP clinical samples, where it achieved an overall specificity and sensitivity of 0.9666 and 0.9722, respectively with a recall of 0.9892 for C_t < 30. Also, the proposed system can be widely used as an accurate, sensitive, rapid, and portable tool to detect COVID-19 in settings where access to a lab is difficult, or the results are urgently required.

Figure 1

(a) Schematic diagram of KU-LAMP showing the dimensions of the device that allows processing of 16 samples simultaneously. The dimensions of the device are 205 mm (l) × 79 mm (w) × 26 mm (h). (b) An exploded view of the device showing the components. The figure is generated in PTC Creo CAD software.

Figure 2

(a) Schematics of the designs used in the thermal modeling for KU-LAMP with the details of material and boundary conditions. The sample compartment block is made from (a) thermally conductive material (i.e., DOWSIL 3-6655 and (b) aluminum with a thin layer of DOWSIL 3-6655 underneath and inside the holes, as a liner. A sample tube is shown for illustrative purposes. The figure is generated in COMSOL Multiphysics software.

Figure 3

(a) The schematics of the model setup in COMSOL Multiphysics with 16 tubes inserted in the block. (b) Temperature distribution in the complete system for t = 0, 1, 5, and 30 min. (c) Temperature profile monitored over 15 min at a point on the upper surface of the block. The figure is generated in COMSOL Multiphysics software.

Figure 4

The block diagram of the proposed system pipeline. The figure is created in MS powerpoint.

Figure 5

Schematics of the steps comprising the protocol for COVID-19 detection using KU-LAMP. (The figure is generated in MS powerpoint).

Figure 6

The proposed COVID-19 screening system. When the input scan is passed to the screening block, it is first pre-processed to remove the background artifacts. Afterward, the preprocessed scan (contain RT-LAMP sample) is passed to the proposed multi-resolution network that extracts discriminative features to differentiate COVID-19 vs. non-COVID-19 subjects. Moreover, the abbreviations within the proposed network are: ZP: Zero-padding, CONV: Convolution, BN: Batch Normalization, ReLU: Rectified Linear Unit, RB: Residual Block, MP: Max pooling, ADD: Addition, CV + ReLU: Convolution layer with ReLU activation, and FC: Fully Connected. (The figure is generated in MS Powerpoint).

Figure 7

Examples of Positive and Negative COVID-19 Scans within the Proposed Dataset.

Figure 8

Performance comparison of proposed network with state-of-the-art classification models in terms of (A) ROC curve, (B) PR curve. (The figure is generated using MATLAB).

Figure 9

Validation results of the RT-LAMP assays performed using KU-LAMP. The photograph shows the pink to yellow/orange color change in the positive samples—with different RT qPCR-derived Ct values — at the end of the 35 min incubation at 65 °C, in comparison to the fuchsia color that is typical for the negative samples (N).

Figure 10

Quantification of the pink-to-yellow/orange color change in the positive samples compared to the Negative Control, as a function of the RT-qPCR cycle threshold C_t. The quantification is performed by computing the color difference between the positive samples and the Negative Control, normalized to the difference between the Positive Control and Negative Control using the R, G, B linear dimensions defining the color space. The standard error bars represent the distribution of the normalized RGB distance for the samples categorized in four Ct bins of 0–25, 25–30, 30–35, and 35–40. (The figure is generated using MATLAB).

Figure 11

The top and side view images of the amplification in the tubes with and without the silicone oil layer. The silicone oil was not added to sample in the tube labeled (“A”), which shows condensation on the sidewalls and the cap. In contrast, the tube labelled as (“B”) contains 5 µl of silicone oil encapsulating the sample and preventing evaporation.