Intentional deep overfit learning for patient-specific dose predictions in adaptive radiotherapy

Abstract

Background: The framework of adaptive radiation therapy (ART) was crafted to address the underlying sources of intra-patient variation that were observed throughout numerous patients' radiation sessions. ART seeks to minimize the consequential dosimetric uncertainty resulting from this daily variation, commonly through treatment planning re-optimization. Re-optimization typically consists of manual evaluation and modification of previously utilized optimization criteria. Ideally, frequent treatment plan adaptation through re-optimization on each day's computed tomography (CT) scan may improve dosimetric accuracy and minimize dose delivered to organs at risk (OARs) as the planning target volume (PTV) changes throughout the course of treatment.

Purpose: Re-optimization in its current form is time-consuming and inefficient. In response to this ART bottleneck, we propose a deep learning based adaptive dose prediction model that utilizes a head and neck (H&N) patient's initial planning data to fine-tune a previously trained population model towards a patient-specific model. Our fine-tuned, patient-specific (FT-PS) model, which is trained using the intentional deep overfit learning (IDOL) method, may enable clinicians and treatment planners to rapidly evaluate relevant dosimetric changes daily and re-optimize accordingly.

Methods: An adaptive population (AP) model was trained using adaptive data from 33 patients. Separately, 10 patients were selected for training FT-PS models. The previously trained AP model was utilized as the base model weights prior to re-initializing model training for each FT-PS model. Ten FT-PS models were separately trained by fine-tuning the previous model weights based on each respective patient's initial treatment plan. From these 10 patients, 26 ART treatment plans were withheld from training as the test dataset for retrospective evaluation of dose prediction performance between the AP and FT-PS models. Each AP and FT-PS dose prediction was compared against the ground truth dose distribution as originally generated during the patient's course of treatment. Mean absolute percent error (MAPE) evaluated the dose differences between a model's prediction and the ground truth.

Results: MAPE was calculated within the 10% isodose volume region of interest for each of the AP and FT-PS models dose predictions and averaged across all test adaptive sessions, yielding 5.759% and 3.747% respectively. MAPE differences were compared between AP and FT-PS models across each test session in a test of statistical significance. The differences were statistically significant in a paired t-test with two-tailed p-value equal to $3.851\times {10}^{-9}$ and 95% confidence interval (CI) equal to [-2.483, -1.542]. Furthermore, MAPE was calculated using each individually segmented structure as an ROI. Nineteen of 24 structures demonstrated statistically significant differences between the AP and FT-PS models.

Conclusion: We utilized the IDOL method to fine-tune a population-based dose prediction model into an adaptive, patient-specific model. The averaged MAPE across the test dataset was 5.759% for the population-based model versus 3.747% for the fine-tuned, patient-specific model, and the difference in MAPE between models was found to be statistically significant. Our work demonstrates the feasibility of patient-specific models in adaptive radiotherapy, and offers unique clinical benefit by utilizing initial planning data that contains the physician's treatment intent.

Keywords: adaptive; artificial intelligence; deep learning; dose prediction; fine-tuning; head and neck cancer; overfit; radiation therapy.

Figure #1:

H&N Patient Data Allocation

Figure #2:

Dose prediction model comparison in terms of MAPE between the adaptive population model and the fine-tuned patient-specific model. Results are averaged across all 26 adaptive treatment plans.

Figure #3:

Dose prediction model comparison in terms of each predicted structure’s max dose as compared to the ground truth, referred to as DMAX error. Results are averaged across all 26 adaptive treatment plans.

Figure #4:

MAPE was calculated using Equation (1) and plotted over time, with time in units of calendar days since the initial CT simulation. The color of each data point represents a unique patient, and data points are connected for patients with greater than 1 adaptive session. Data points shaped as a circle represent the performance of the population model for that specific adaptive session and may be connected by dotted lines. Data points shaped as a star represent the performance of the fine-tuned patient-specific model and may be connected by a solid line.

Figure #5:

Dose displays, shown for a session from an individual patient. The top panel displays the ground truth dose, while the middle and bottom panels display dose as predicted by the AP and FT-PS models respectively.

Figure #6:

Dose displays, shown for the same patient displayed in figure #5. The top panel displays the ground truth dose. The middle and bottom panel display dose differences in terms of percent error, where a positive value indicates over-prediction of dose and a negative value indicates under-prediction of dose.

Figure #7:

The dose volume histogram for the patient displayed in figure #5 and figure #6. The solid lines display the ground truth dose delivered to each OAR, whereas dotted lines represent the AP model’s prediction for each OAR and dashed lines represent the FTPS model’s prediction for each OAR.