Female pelvic synthetic CT generation based on joint intensity and shape analysis

Area of Science:

• Medical imaging
• Radiotherapy
• Image processing

Background:

• Magnetic Resonance Imaging (MRI) offers superior soft tissue contrast compared to Computed Tomography (CT) for radiotherapy planning.
• Aligning MRI and CT images can introduce systematic errors.
• Developing MRI-only workflows is desirable for improved accuracy and efficiency in radiotherapy.

Purpose of the Study:

• To present a method for generating synthetic CT (sCT) volumes from MRI data.
• To address challenges in MRI-based sCT generation, particularly tissue intensity overlap.
• To evaluate the accuracy and potential of MRI-only radiotherapy treatment planning.

Main Methods:

• Probabilistic tissue classification of voxels from T1 Dixon MRI sequences.
• Integration of pelvic bone shape information using principal component analysis to enhance intensity-based classification.
• Application of a regularization term within a fuzzy c-means classification scheme.
• Generation of sCT images by combining class probabilities with assigned attenuation values.

Main Results:

• Generated sCT images exhibited intensity distributions similar to CT images.
• Mean absolute errors ranged from 13.7 HU (muscle) to 274.4 HU (bone).
• Radiotherapy plans optimized with sCT showed minimal dose differences compared to CT-based plans (e.g., D0.5cc mean difference 0.21 Gy).

Conclusions:

• The proposed method accurately generates synthetic CT images from MRI data.
• The integration of bone shape information effectively overcomes challenges in tissue classification.
• This MRI-only approach shows significant potential for improving radiotherapy treatment planning and image guidance.