Clinical implications of dose to functional lung volumes in the trimodality treatment of esophageal cancer.

BACKGROUND: Trimodality treatment, i.e., neoadjuvant chemoradiotherapy (nCRT) followed by surgery, for locally advanced esophageal cancer (EC) improves overall survival but also increases the risk of postoperative pulmonary complications. Here, we tried to identify a relation between dose to functional lung volumes (FLV) as determined by 4D-CT scans in EC patients and treatment-related lung toxicity. MATERIALS AND METHODS: All patients with EC undergoing trimodality treatment between 2017 and 2022 in UZ Leuven and scanned with 4D-CT-simulation were selected. FLVs were determined based on Jacobian determinants of deformable image registration between maximum inspiration and expiration phases. Dose/volume parameters of the anatomical lung volume (ALV) and FLV were compared between patients with versus without postoperative pulmonary complications. Results of pre- and post-nCRT pulmonary function tests (PFTs) were collected and compared in relation to radiation dose. RESULTS: Twelve out of 51 EC patients developed postoperative pulmonary complications. ALV was smaller while FLV10Gy and FLV20Gy were larger in patients with complications (respectively 3141 ± 858mL vs 3601 ± 635mL, p = 0.025; 360 ± 216mL vs 264 ± 139mL, p = 0.038; 166 ± 106mL vs 118 ± 63mL, p = 0.030). No differences in ALV dose-volume parameters were detected. Baseline FEV1 and TLC were significantly lower in patients with complications (respectively 90 ± 17%pred vs 102 ± 20%pred, p = 0.033 and 93 ± 17%pred vs 110 ± 13%pred, p = 0.001), though no other PFTs were significantly different between both groups. DLCO was the only PFT that had a meaningful decrease after nCRT (85 ± 17%pred vs 68 ± 15%pred, p < 0.001) but was not related to dose to ALV/FLV. CONCLUSION: Small ALV and increasing FLV exposed to intermediate (10 to 20 Gy) dose are associated to postoperative pulmonary complications. Changes of DLCO occur during nCRT but do not seem to be related to radiation dose to ALV or FLV. This information could attribute towards toxicity risk prediction and reduction strategies for EC.

Automated clinical decision support system with deep learning dose prediction and NTCP models to evaluate treatment complications in patients with esophageal cancer.

BACKGROUND AND PURPOSE: This study aims to investigate how accurate our deep learning (DL) dose prediction models for intensity modulated radiotherapy (IMRT) and pencil beam scanning (PBS) treatments, when chained with normal tissue complication probability (NTCP) models, are at identifying esophageal cancer patients who are at high risk of toxicity and should be switched to proton therapy (PT). MATERIALS AND METHODS: Two U-Net were created, for photon (XT) and proton (PT) plans, respectively. To estimate the dose distribution for each patient, they were trained on a database of 40 uniformly planned patients using cross validation and a circulating test set. These models were combined with a NTCP model for postoperative pulmonary complications. The NTCP model used the mean lung dose, age, histology type, and body mass index as predicting variables. The treatment choice is then done by using a ΔNTCP threshold between XT and PT plans. Patients with ΔNTCP ≥ 10% were referred to PT. RESULTS: Our DL models succeed in predicting dose distributions with a mean error on the mean dose to the lungs (MLD) of 1.14 ± 0.93% for XT and 0.66 ± 0.48% for PT. The complete automated workflow (DL chained with NTCP) achieved 100% accuracy in patient referral. The average residual (ΔNTCP ground truth - ΔNTCP predicted) is 1.43 ± 1.49%. CONCLUSION: This study evaluates our DL dose prediction models in a broader patient referral context and demonstrates their ability to support clinical decisions.