Vision 20/20: perspectives on automated image segmentation for radiotherapy.

Area of Science:

• Medical physics
• Radiotherapy
• Medical image analysis

Background:

• Manual segmentation of medical images for radiation therapy (RT) is standard but time-consuming and variable.
• Advances in image guidance and treatment adaptation in RT necessitate faster, more accurate segmentation.
• Automated segmentation (autosegmentation) aims to reduce workload and standardize organ delineation.

Purpose of the Study:

• To review current autosegmentation methods for RT applications.
• To outline the strengths and limitations of these methods.
• To propose strategies for wider clinical adoption of autosegmentation.

Main Methods:

• Review of existing literature on autosegmentation techniques in radiation therapy.
• Analysis of methods relevant to image guidance and treatment adaptation.
• Evaluation of computational requirements and processing times.

Main Results:

• Autosegmentation provides an efficient tool for RT planning, offering a valuable starting point for clinicians.
• Modern hardware, including GPUs, enables autosegmentation tasks to be completed within minutes.
• Current limitations exist, but improvements are expected through standardization and multimodality approaches.

Conclusions:

• Autosegmentation is a valuable tool for RT planning, enhancing efficiency and consistency.
• Near-term improvements will focus on CT-based tools and protocol standardization.
• Long-term advancements anticipate multimodality imaging and integration of biological/pathological data.