Published online Feb 24, 2021. doi: 10.5306/wjco.v12.i2.95
Peer-review started: November 20, 2020
First decision: December 3, 2020
Revised: December 7, 2020
Accepted: December 22, 2020
Article in press: December 22, 2020
Published online: February 24, 2021
Cardiotoxicity from thoracic radiotherapy can significantly contribute to treatment related morbidity and mortality. Despite this, cardiac substructures are not routinely delineated in thoracic radiation planning.
Autosegmentation of cardiac substructures would allow for relative dose calculation to these subsites without the added labor of manual definition.
To determine whether autosegmentation software can be successfully employed for the cardiac substructures in patients planned using respiratory gated, non-contrasted computed tomography (CT) imaging.
This retrospective study included patients who underwent stereotactic body radiation therapy (SBRT) for inoperable, early-stage non-small cell lung cancer from 2007 to 2019. All patients were simulated via CT imaging with respiratory gating without intravenous contrast. A 20-patient atlas of the cardiac substructures was manually constructed and used to facilitate autosegmentation via MIM software. A total of three iterations of autosegmentations were completed, each using 10 patients. Generated structure quality was evaluated by degree of required manual edits and volume discrepancy between the autocontoured structures and its edited sister structure.
The great vessels and heart chambers were reliably autosegmented with most edits considered minor. In contrast, coronary arteries either failed to be autosegmented or the generated structures required major alterations necessitating deletion and manual definition. Similarly, the generated mitral and tricuspid valves were poor whereas the aortic and pulmonary valves required at least minor and moderate changes respectively. For the majority of subsites, the additional samples did not appear to substantially impact the quality of generated structures. Volumetric analysis between autosegmented and its manually edited sister structure yielded comparable findings to the physician-based assessment of structure quality.
Our study indicated that the great vessels and heart chambers can be reliable autocontoured using MIM software. On the other hand, autosegmentation for valves is inconsistent and poor for coronary arteries. Anatomic variances and/or implanted hardware may impact the quality of autosegmentation.
Radiation heart dose is an important dosimetric parameter however dose tolerances for the cardiac substructures in conventional therapy and SBRT are not well-established. Therefore, artificial intelligence based contouring programs allow dose to be calculated to the select cardiac subsites without the added labor of manual definition.