Abstract
Purpose: To evaluate the clinical need for automated decision-support platforms for Adaptive Radiotherapy Therapy (ART) of Head & Neck cancer (HNC) patients. Methods: We tested RTapp (SegAna), a new decision-support software for ART, to investigate 22 HNC patients' data retrospectively. For each fraction, RTapp estimated the daily and cumulative doses received by targets and organs at risk (OARs) from daily 3D imaging in real-time. RTapp also included a prediction algorithm that analyzed dosimetric parameters (DP) trends against dosimetric endpoints (DE) to trigger adaptation up to 4 fractions ahead. Results: RTapp addressed the needs of treatment adaptation. 15/22 studies (68%) violated PTV coverage or parotid mean dose ($D_{mean}$) at the end of treatment (EOT). 15 parotids were flagged for exceeding $D_{mean}$ constraints with a median increase of +3.18 Gy. The software predicted violations up to 4 fractions ahead with high accuracy. Conclusion: Integrated into an ART clinical workflow, RTapp can predict whether specific treatment would require adaptation ahead of time, potentially streamlining clinical decision-making.
Introduction
Head and neck cancer (HNC) patients undergo significant anatomical changes during their 6–7 weeks of radiation treatment, including weight loss, tumor shrinkage, and internal organ shifts. Because modern radiation is highly conformal, even small shifts can lead to "cold spots" in the tumor or "hot spots" in healthy organs like the parotid (saliva) glands. While Adaptive Radiotherapy (ART)—the process of adjusting the plan mid-treatment—exists, it is currently a clinical bottleneck. "Offline" ART typically requires hours of manual work, often leading to a 2–5 day delay. Clinicians need automated tools to flag which patients require a replan and, ideally, predict that need before dose limits are exceeded.
Methodology and Workflow
This study retrospectively analyzed data from 22 HNC patients treated at Loyola University Chicago. The cohort primarily included pharynx cases (15), along with oral cavity, larynx, and other sites. The evaluation focused on RTapp (SegAna), a decision-support platform designed to automate the ART workflow through three primary steps:
- Deformable Image Registration (DIR): An optical flow-based algorithm maps the initial planning CT to daily Cone Beam CT (CBCT) images.
- Dose Mapping: The software uses the resulting deformation vector field (DVF) to "warp" the original dose distribution onto the patient's daily anatomy.
- Trend Analysis & Prediction: RTapp calculates daily and cumulative Dose Volume Histograms (DVHs) and uses a linear regression model to forecast whether a patient will hit a "warning" or "adaptation" threshold within the next four fractions.
Specific thresholds were defined: a "Warning" if PTV coverage fell below 95% ($V_{95} < 95\%$) and an "Adaptation" trigger if it fell below 93%. For OARs, a 10% increase in mean or max dose served as the trigger.
Key Results and Clinical Insights
The retrospective analysis revealed a high clinical need for adaptation, with 68% (15/22) of patients flagged for violating either PTV coverage or parotid dose limits by the EOT. For flagged PTVs, the mean coverage decrease was -7.7%. For parotids, the median dose increase was +3.18 Gy.
Several critical insights emerged from the data:
- The "Replanning Gap": By predicting violations up to four fractions ahead, RTapp effectively cancels out the administrative delay required to create a new treatment plan, ensuring the new plan is ready exactly when needed.
- Unpredictability of OARs: Unlike tumor coverage, which showed some clustering in timing, the need for parotid gland adaptation showed no specific temporal trend. This suggests that OAR sparing must be monitored every single fraction.
- Prediction Accuracy: The model was highly accurate, with a mean difference between predicted and calculated values of only -2.9% for PTV coverage and 0.2 Gy for parotid mean dose. Accuracy improved as the software moved closer to the actual threshold-crossing fraction.
Discussion and Conclusion
RTapp demonstrates the potential to transform ART from a reactive, manual chore into a proactive, automated workflow. While the current study is limited by its sample size and the use of linear regression for inherently non-linear anatomical changes, the results are promising. Future work will involve larger cohorts and the potential integration of machine learning models to further improve forecasting. Ultimately, integrating such decision-support tools into the clinical workflow could significantly reduce the risk of under-treating tumors or over-irradiating healthy tissue in HNC patients.
Summary updated