An accurate algorithm to match imperfectly matched images for lung tumor detection without markers

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Anatomy

Background:

• Accurate lung tumor localization is crucial for effective radiotherapy.
• Respiration-induced tumor motion complicates image-guided interventions.
• Existing methods struggle with image discrepancies and lack of internal markers.

Purpose of the Study:

• To develop and validate a novel algorithm for precise lung tumor localization on kV projection images.
• To overcome challenges posed by respiratory motion and inter-image intensity variations.
• To enable markerless tumor tracking for improved radiotherapy accuracy.

Main Methods:

• Digitally Reconstructed Radiographs (DRRs) were generated from 3D CT data.
• A two-step matching algorithm was employed: background matching followed by tumor localization.
• Normalized Cross-Correlation (NCC) and pixel-based linear transformation were used for image registration.
• Tumor regions were isolated by subtracting matched background DRRs from projection images.

Main Results:

• The algorithm successfully located tumors on kV fluoroscopy images without internal markers.
• Dynamic tumor tracking was achieved on phantom studies with high accuracy.
• Maximum localization error was < 2.2 mm, with an average error < 0.9 mm across 12 gantry angles.
• Robust performance was demonstrated despite strong background signals from surrounding anatomy.

Conclusions:

• The developed algorithm offers a simple yet efficient solution for markerless lung tumor localization.
• It effectively handles respiratory motion and image artifacts, enhancing precision in image-guided radiotherapy.
• This technique holds significant potential for improving treatment accuracy and patient outcomes in lung cancer therapy.