Characterisation of radiotherapy planning volumes using textural analysis

Area of Science:

• Medical imaging analysis
• Radiotherapy
• Artificial intelligence in oncology

Background:

• Current artificial intelligence methods for gross tumour volume (GTV) delineation on CT and MR images lack the accuracy needed for radiotherapy.
• Accurate GTV definition is crucial for effective radiotherapy planning and delivery.

Purpose of the Study:

• To develop and evaluate an image analysis method for classifying distinct regions within the GTV and other relevant areas on CT images.
• To assess the potential of this method for improving radiotherapy applications, particularly automated outlining.

Main Methods:

• An image analysis method was applied to CT images from eight bladder cancer patients at planning and during treatment.
• Statistical and fractal textural features (N=27) were calculated for the bladder, rectum, and control regions across axial, coronal, and sagittal planes.
• Unsupervised classification was performed using a reduced feature set (N=3).

Main Results:

• The developed image analysis method demonstrated significant classification accuracy on axial, coronal, and sagittal CT image planes.
• A reduced feature set of three features proved sufficient for accurate classification.
• The approach shows promise for further development in radiotherapy applications.

Conclusions:

• The proposed image analysis method offers a viable approach for classifying clinically relevant regions in CT images for bladder cancer radiotherapy.
• This technique has the potential to be advanced towards an automatic outlining system, enhancing radiotherapy precision.
• Further development is warranted to fully integrate this method into clinical radiotherapy workflows.