Platform Module 01
AI-Assisted Beam-Segment Candidate Generation
Airato's Auto-Planning module generates initial beam configurations from CT anatomy and learned planning patterns, reducing physicist setup time for standard IMRT treatment sites.
How the Auto-Planning Engine Works
The Auto-Planning module takes the CT image set, OAR contours, and PTV definition as inputs. Using a convolutional neural network trained on historical IMRT cases, the system identifies anatomical feature patterns and maps them to beam configurations that historically achieved acceptable DVH outcomes.
Output is a set of 5–7 beam-angle candidates with initial fluence maps. These are presented to the physicist as starting points — the physicist selects, refines, or discards any candidates before proceeding to dose optimization.
The system does not make autonomous treatment decisions. Every output requires explicit physicist acceptance before entering the optimization phase.
Algorithm Overview
Anatomy-Aware Feature Extraction
CNN processes CT slices to identify PTV geometry, OAR proximity patterns, and body habitus features relevant to beam path selection.
Historical Pattern Matching
Beam angles correlated with historically acceptable DVH outcomes for similar anatomical configurations. Trained on real-world cases reviewed by medical physicists.
Physicist Override at Every Step
All candidate configurations are editable. Beam angles, segment weights, and MLC positions can be manually adjusted before optimization proceeds.
Supported IMRT Planning Sites
Auto-Planning has been characterized on standard IMRT treatment sites. Performance varies by site complexity — see dosimetric benchmarks for validation cohort details.
| Treatment Site | Standard Beam Count | Auto-Planning Status |
|---|---|---|
| Prostate | 7–9 beams | Characterized |
| Head & Neck | 7–9 beams | Characterized |
| Lung | 5–7 beams | Characterized |
| Cervix | 5–7 beams | In evaluation |
| Rectum | 5–7 beams | In evaluation |
Related Platform Modules
Dose Optimization
DVH-constraint solver for multi-criteria OAR sparing and target coverage. Works with Auto-Planning output or manual beam configurations.
Learn moreAdaptive RT
Daily re-planning module for anatomy-adaptive treatment delivery. Integrates with the same plan generated by Auto-Planning.
Learn moreClinical Evidence
Dosimetric benchmarks and validation data for Auto-Planning's beam configuration performance across characterized treatment sites.
View benchmarksSee Auto-Planning with a representative case.
Our clinical team can walk through a prostate or head-and-neck case demonstrating the full candidate generation cycle.