Sorting lung tumor volumes from 4D-MRI data using an automatic tumor-based signal reduces stitching artifacts.

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• 4D-magnetic resonance imaging (4D-MRI) is crucial for visualizing lung tumors during respiration.
• Respiratory motion introduces artifacts in 4D-MRI, complicating tumor volume assessment.
• Current sorting methods rely on anatomical signals, which may not accurately capture tumor motion.

Purpose of the Study:

• To evaluate a novel signal derived from tumor motion for improved 4D-MRI data sorting.
• To compare the efficacy of a tumor motion signal against anatomical signals in reducing stitching artifacts.
• To enhance the precision of lung tumor volume reconstruction in 4D-MRI.

Main Methods:

• Collected 4D-MRI scans from 10 lung cancer patients.
• Generated a tumor-motion signal (TumorPC1) using the first principal component of tumor neighborhood movement.
• Compared TumorPC1 with anatomical signals (diaphragm, chest wall, body area) using Pearson correlation and a roughness metric (Rg) to assess stitching artifacts.

Main Results:

• The TumorPC1 signal showed the strongest correlation with superior-inferior tumor motion (median r=0.86).
• Image stacks sorted using TumorPC1 exhibited significantly lower roughness (fewer stitching artifacts) compared to those sorted by anatomical signals (p=0.02-0.05).
• TumorPC1 significantly outperformed chest wall and body area signals in correlation with tumor motion (p<0.05).

Conclusions:

• A novel signal derived from tumor motion (TumorPC1) enables more precise sorting of 4D-MRI data.
• This approach significantly reduces stitching artifacts, leading to smoother 3D tumor volume reconstructions.
• Tumor motion-based sorting offers a superior alternative to traditional anatomical signal-based methods for lung tumor imaging.