Statistical deformation reconstruction using multi-organ shape features for pancreatic cancer localization

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiotherapy Physics

Background:

• Respiratory motion significantly deforms abdominal organs and tumors, complicating clinical applications.
• While single organ deformations are well-studied, inter- and intra-patient multi-organ deformations remain statistically unformulated.
• Accurate modeling of multi-organ deformations is crucial for effective adaptive radiotherapy.

Purpose of the Study:

• To develop a statistical multi-organ deformation library for abdominal organs.
• To apply this library for deformation reconstruction using organ shape features.
• To predict pancreatic cancer displacement for adaptive radiotherapy using a per-region learning approach.

Main Methods:

• Generated statistical multi-organ motion/deformation models for stomach, liver, kidneys, and duodenum.
• Utilized shape matching on region labels from 4D computed tomography (CT) images of 25 pancreatic cancer patients (250 volumes).
• Employed a per-region-based deformation learning with a non-linear kernel model to predict organ displacement.

Main Results:

• The proposed multi-organ deformation library demonstrated superior performance compared to general per-patient models.
• Achieved clinically acceptable estimation errors: mean distance of 1.2 ± 0.7 mm and Hausdorff distance of 4.2 ± 2.3 mm.
• Accurate deformation estimation was maintained throughout respiratory motion.

Conclusions:

• The developed multi-organ deformation library enables statistically formulated modeling of complex abdominal organ motion.
• The per-region learning approach effectively predicts organ displacement for adaptive radiotherapy.
• This method offers a significant advancement in precision and accuracy for radiotherapy planning.