JCS

JCS performs joint classification and segmentation of chest CT images to diagnose COVID-19 and localize pulmonary opacifications.

Key Features:

• Real-time Diagnosis: Reduces per-case processing time for chest CT scans relative to traditional methods, with an approximate reported processing time of 21.5 minutes.
• Explainability: Produces explainable outputs that provide insights into the model's decision-making process.
• High Sensitivity and Specificity: Achieves an average sensitivity of 95.0% and specificity of 93.0% for COVID-19 classification.
• Accurate Segmentation: Attains a Dice score of 78.5% for segmentation of opacifications in lung parenchyma.

Scientific Applications:

• Complementary diagnostic to RT-PCR: Serves as a complementary modality to Reverse Transcription Polymerase Chain Reaction (RT-PCR) by detecting COVID-19-related findings on chest CT scans.
• Early-stage detection: Supports identification of early-stage COVID-19 patients through detection and delineation of pulmonary opacifications on chest CT.

Methodology:

Developed using the COVID-CS dataset of 144,167 chest CT images from 400 COVID-19 patients and 350 uninfected cases, with 3,855 images from 200 patients annotated with fine-grained pixel-level opacification labels plus lesion counts, areas, and locations, and implemented via a joint classification and segmentation approach on chest CT images.