Related Experiment Video
Updated: Apr 29, 2026
06:21
Adaptable Angled Stereotactic Approach for Versatile Neuroscience Techniques
Published on: May 7, 2020
4.8K
Beam orientation in stereotactic radiosurgery using an artificial neural network
Agnieszka Skrobala1, Julian Malicki1
1Department of Electroradiology, University of Medical Science, Poznan, Poland; Department of Medical Physics, Greater Poland Cancer Centre, Poznan, Poland.
Summary
Artificial neural networks (ANNs) can generate beam orientations for stereotactic radiosurgery (SRS). ANN-generated plans match human-designed plans, showing feasibility for automated treatment planning.
Area of Science:
- Medical Physics
- Artificial Intelligence in Medicine
- Radiotherapy
Background:
- Stereotactic radiosurgery (SRS) requires precise beam orientation for effective treatment.
- Automating treatment planning can improve efficiency and consistency.
Purpose of the Study:
- To assess the feasibility of using artificial neural networks (ANNs) to generate beam orientations in SRS.
- To compare ANN-generated treatment plans with those created by human planners.
Main Methods:
- Three ANNs were developed and trained using a dataset of 669 intracranial lesions.
- Different input methods (Cartesian coordinates, genetic algorithms, vectors) were explored for ANN models.
- ANN-generated plans were evaluated against human-generated plans using dose-volume histograms, RMS error, and Gamma index.
Main Results:
- ANN models successfully generated beam orientations for SRS treatment plans.
- No significant differences were found in target coverage (95% isodose) between ANN and human plans.
- ANN1, using Cartesian coordinates, demonstrated the best agreement with human planners based on Gamma index and RMS error.
Conclusions:
- ANNs are a feasible tool for determining beam orientation in stereotactic radiosurgery.
- Treatment plans generated by ANNs are comparable in quality to those designed by human experts.
