UM - Automated detection and segmentation NSCLC

UM - Automated detection and segmentation NSCLC performs automated detection and volumetric segmentation of non-small cell lung cancer (NSCLC) on computed tomography (CT) images to enable quantitative tumor measurement and prognostic assessment.

Key Features:

• Fully Automated Pipeline: Performs automated detection and volumetric segmentation of NSCLC on CT images without manual intervention.
• Comprehensive Validation: Validated across 1,328 thoracic CT scans from eight institutions encompassing varying image slice thicknesses, tumor sizes, levels of interpretation difficulty, and tumor locations.
• Performance Metrics: Provides quantitative performance metrics assessed across slice thickness, tumor size, interpretation difficulty, and tumor location.
• Clinical Trial Evaluation: Evaluated in an in-silico prospective clinical trial where automatic segmentations outperformed expert manual segmentation in speed and reproducibility and were preferred by radiologists and radiation oncologists in 56% of cases.
• Prognostic Evaluation: Uses tumor volumes derived from automated contours with RECIST criteria to stratify patients into low and high survival groups, showing higher statistical significance than manual contouring.

Scientific Applications:

• Diagnosis and Patient Management: Provides accurate detection and segmentation to support diagnostic assessment and clinical decision making for NSCLC.
• Radiotherapy Planning: Supplies precise volumetric tumor measurements to inform targeted radiotherapy planning.
• Response Evaluation: Enables consistent longitudinal monitoring of tumor size and volumetric changes for treatment response assessment.
• Quantitative Image Research: Facilitates quantitative analysis and comparison across large multicenter CT datasets for imaging research.

Methodology:

Automated detection and volumetric segmentation on CT images with tumor volume measurement from automated contours; validation performed on 1,328 thoracic CT scans from eight institutions and evaluated in an in-silico prospective clinical trial.