Smartphone-based device for point-of-care diagnostics of pulmonary inflammation using convolutional neural networks (CNNs)

Abstract

In pulmonary inflammation diseases, like COVID-19, lung involvement and inflammation determine the treatment regime. Respiratory inflammation is typically arisen due to the cytokine storm and the leakage of the vessels for immune cells recruitment. Currently, such a situation is detected by the clinical judgment of a specialist or precisely by a chest CT scan. However, the lack of accessibility to the CT machines in many poor medical centers as well as its expensive service, demands more accessible methods for fast and cheap detection of lung inflammation. Here, we have introduced a novel method for tracing the inflammation and lung involvement in patients with pulmonary inflammation, such as COVID-19, by a simple electrolyte detection in their sputum samples. The presence of the electrolyte in the sputum sample results in the fern-like structures after air-drying. These fern patterns are different in the CT positive and negative cases that are detected by an AI application on a smartphone and using a low-cost and portable mini-microscope. Evaluating 160 patient-derived sputum sample images, this method demonstrated an interesting accuracy of 95%, as confirmed by CT-scan results. This finding suggests that the method has the potential to serve as a promising and reliable approach for recognizing lung inflammatory diseases, such as COVID-19.

Figure 1

During the inflammation phase of COVID-19, numerous immune cells are mobilized into the lung environment through the vasodilation process. The vessels surrounding the alveoli undergo increased permeability, allowing the entry of blood contents into the lung environment. This phenomenon has the potential to alter the concentration of sputum components, including electrolyte salts.

Figure 2

(A) CT scan from the two patients with COVID-19 disease and with and without lung inflammation. Signs such as glass ground opacification (GGO) in the CT images imply lung involvement. (B–D) Display the results of the Na, K, and NA + K concentrations in the negative and positive CT cases, respectively. (E) Workflow of the designed method for analyzing the fern structures in the dried sputum samples of the patients with different CT scan results. (F) Different ferning patterns in the air-dried samples of the CT positive and negative cases. (G) The percentage of the fern area in the sputum images of the CT positive and negative cases.

Figure 3

(A) Sample slide for dropping the sputum sample (B) Mini-microscope system for visualization and imaging of the air-dried sputum sample. (C) The main device and closed view (C-1) of the mini-microscope. (D) Exploded view of the designed mini-microscope for sputum analysis. (E) Placing the smartphone on the mini-microscope and imaging the sputum sample as well as its analysis by the (F) AI-based application.

Figure 4

System performance. (A) Computational flow of data in the utilized neural network, for transfer learning, the pre-trained EfficientNet was retrained by using our dataset of 650 salivary images derived from 70 participants. (B) EfficientNet-B0 structure; this mobile-sized architecture contains 7 main blocks, each containing a varying number of sub-blocks. (C) Classifier layers added for retraining and decide about the ferning patterns. (D,E) training and validation curves for accuracy and cross-entropy of the network; after 80 epochs model achieved a validation accuracy of 98.23% on training set and the validation cross-entropy was 0.18. (F) To evaluate the diagnostic ability of the system, receiving operative characteristic (ROC) curve was plotted for different thresholds. The area under the ROC curve (AUC) showed a value of 0.99. (G) the confusion matrices for the test sets when smartphone-based device analyzed the patient-derived samples. True classes are determined by CT-scan results.