Related Experiment Video
Updated: Jan 11, 2026
08:34
Proton Therapy Delivery and Its Clinical Application in Select Solid Tumor Malignancies
Published on: February 6, 2019
20.9K
Non-convex IMRT treatment planning using deep inverse optimization
Yang Lei1, Jiahan Zhang1, Tian Liu1
1Department of Radiation Oncology, Icahn School of Medicine at Mount Sinai, New York, New York, USA.
Medical Physics
|November 13, 2025
Summary
This study introduces a novel deep inverse optimization (DIO) method for intensity-modulated radiation therapy (IMRT) planning. The approach efficiently handles complex dose constraints, improving treatment plan quality and significantly reducing optimization time for lung cancer patients.
Area of Science:
- Radiation Oncology
- Medical Physics
- Computational Biology
Background:
- Intensity-modulated radiation therapy (IMRT) uses inverse optimization for precise dose delivery.
- Non-convex dose-volume constraints pose significant challenges in IMRT planning due to complexity.
- Optimizing IMRT plans requires balancing tumor coverage with organ-at-risk (OAR) sparing.
Purpose of the Study:
- To develop and evaluate a deep inverse optimization (DIO) approach for IMRT fluence map optimization (FMO).
- To approximate non-convex dose-volume constraints efficiently while maintaining convexity.
- To improve adherence to clinical dose objectives in IMRT planning.
Main Methods:
- Introduced a novel relaxation technique using second-order cone constraints to approximate non-convex conditions.
- Employed a DIO framework to solve the resulting second-order cone programming problem for treatment plan generation.
- Utilized sequential optimization to balance tumor coverage and OAR sparing in 30 locally advanced non-small cell lung cancer (NSCLC) patient cases.
Main Results:
- Achieved comparable planning target volume (PTV) coverage with improved dose homogeneity and reduced hotspots.
- Demonstrated significant OAR sparing, including reduced lung V5Gy, heart V30Gy, and spinal cord maximum dose.
- Showcased significantly faster optimization convergence (approximately 5 iterations vs. 39).
Conclusions:
- The proposed DIO method effectively approximates non-convex dose-volume constraints using second-order cone programming.
- This approach enhances IMRT plan quality through improved dose homogeneity and OAR sparing.
- The method offers a computationally efficient and practical solution for advanced radiation therapy planning, accelerating optimization convergence.
