Detecting COVID-19 from chest computed tomography scans using AI-driven android application.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic significantly impacted global health, straining healthcare systems worldwide.
• Chest Computed Tomography (CT) scans are crucial for diagnosing and predicting COVID-19 outcomes.
• There is a need for accessible, cost-efficient diagnostic tools for COVID-19 on resource-limited devices.

Purpose of the Study:

• To develop a novel Android application for detecting COVID-19 infection from chest CT scans.
• To implement a deep learning algorithm for high accuracy and efficiency in diagnosis.
• To create an attention heatmap visualizing lung infection regions for improved clinical interpretation.

Main Methods:

• Development of a deep learning algorithm integrated into an Android application for COVID-19 detection from chest CT scans.
• Implementation of an attention heatmap algorithm to highlight infected lung parenchyma regions.
• Utilization of a novel selection approach combined with multi-threading to accelerate heatmap generation on mobile devices, reducing processing time by 93%.

Main Results:

• The deep learning model achieved high performance metrics: 99.58% F1 score, 99.58% accuracy, and 99.69% sensitivity.
• The attention heatmap effectively visualized infection regions within the lung parenchyma, verified by radiologists.
• The mobile application demonstrated significantly reduced processing times for heatmap generation.

Conclusions:

• The developed Android application offers a swift, mobile, and accessible solution for COVID-19 diagnosis using chest CT scans.
• The high accuracy and efficiency of the deep learning algorithm and heatmap visualization aid in early diagnosis and patient triage.
• This technology has the potential to benefit high-volume clinical practices by enabling quick and efficient patient assessment.