The BERA method offers the opportunity for virus detection in minutes, without the use of any extraction protocol. To achieve this, the matrix effect of various VTM should be addressed (Sun, 2018). After testing a series of VTM, the different responses that were observed on their potential difference (millivolts) were analyzed. A specific algorithms were then developed for each VTM to assess the detection of SARS-CoV-2.

Data were uploaded on the online database using Google Firestore and Google Cloud Functions to run analytics. A specific algorithm which was developed according to a previously described procedure (Hadjilouka et al., 2020) was used to produce/calculate the final results.

There are four different stages prior to result analysis:

Data set: Contained measurements from negative and positive samples. Each measurement consisted of a time series of potentiometric measurements (in Volts).

Training/test data set: The data set was divided into training and test data sets. The training data set was used to determine the algorithm limits and the test data set was used to evaluate the algorithm.

Editing / Exporting features: The data set was processed in a two-step process. In the first step, the background noise was subtracted to normalize and calibrate the signal and in the second step, the purified data was used as input for the development of an algorithm capable of detecting positive and negative samples. Each feature vector was calculated based on (a) the average values (Mean) for each cleaned data set, (b) the rolling average with rolling window size 50 (Min Sums), and (C) the rolling average with rolling window size 50 (Max Diffs), as it was described by Hadjilouka et al. (2020). This procedure was applied in each electrode channel and the overall test data set.

Algorithm: The algorithm used the feature vectors from the previous step as input for generating/calculating the final results. Three thresholds were set for the mean values, the minimum sums, and the maximum differences, and were compared with the corresponding values from each measurement. The final result was the dominant result (e.g. if 6 electrodes had a ‘Positive’ result and 2 had a ‘Negative’ result based on the thresholds, the result would be ‘Positive’) obtained from the above calculations.

The result is displayed on smartphone after being compared to the 3 different thresholds obtained from the data analysis. The tests offer an easy-to-use user interface (UI) with simple login option parameters such as temperature of person under testing and it runs directly. The process of the data analysis is summarized in Fig. 2 .

Data analysis process on cloud functions. The analysis is being done in real time after the tests completion (Figure adapted from Hadjilouka et al. (2020). Newly Developed System for the Robust Detection of Listeria monocytogenes Based on a Bioelectric Cell Biosensor. Biosensors (Basel), Nov 17;10(11):178).

From the experiments, we observed a biosensor response due to the enhanced matrix effect of the Viral Transport Media (VTM). Consequently, further research to identify and eradicate possible impediments due to the matrix effect is required for the optimization of the system. On the other hand, utilizing the receiver operating characteristic (ROC) analysis was possible to define the optimal model for increasing the specificity of the system, having in mind that this will decrease the sensitivity of the developed system.

A ROC curve was designed (not shown) in order to illustrate the diagnostic ability of the system as its discrimination threshold was varied. To draw a ROC curve, only the true-positive rate (TPR) and the false-positive rate (FPR) were needed. A ROC space was defined by FPR and TPR as x and y axes, respectively, which depicted relative trade-offs between true positive and false positive.