Session Item

Recently, Sentker et al reported that artifacts in 4DCT due to irregular breathing and under-sampling are associated with local recurrence in 62 lung and liver SABR patients (R&O 2020). This study suggested that acquisition quality of 4DCT could have a measurable impact on treatment quality and survival. The aim of this study is to investigate this finding in a large cohort of lung cancer patients.

For 991 lung cancer patients treated with conventional radiotherapy between 2017 and 2020, DICOM objects of breathing traces were collected from a pressure belt during 4DCT. In our clinical protocol, scans of slow breathing patients can be too short to cover the entire lungs, because scan time is limited at 120 s. In this case a free breathing scan is also taken. In-house software was used to collect the total scan duration, average and standard deviation (SD) of breathing period (using peak detection after Gaussian smoothing), and average and SD of belt amplitude. Correlations between the breathing trace parameters and clinical variables including age, tumour stage, performance status (PS), comorbidity ACE score, and tumour volume were determined. Univariable and multivariable Cox survival models were constructed with and without breathing trace parameters.

Visual inspection of a subset of scans showed moderate artifacts in ~25% of scans, but very rarely at the tumour level. The average breathing period was negatively associated with tumour volume (Wilcoxon test at median period, p=0.003) and PS (p=0.02) but was strongly correlated with total scan time (r=0.63). In univariable Cox analysis (Table I), a longer average breathing period was associated with longer survival (HR=0.89 s^-1, p=0.026) as demonstrated in Fig 1a. Also, a longer scan time was associated with longer survival, especially for scan times close to 120 s (Fig. 1b). None of the parameters describing breathing irregularity, as proxy for 4DCT artifacts, were associated with survival in univariable analysis. In multivariable modeling (Table I), combining commonly used clinical variables (A) with breathing period (B) did not improve the model.  When combining clinical variables with total scan time (C), total scan time > 119 s was significant for worse outcome (p=0.004).

This study did not find that 4DCT artifacts from breathing irregularity are associated with poorer survival. Both longer scan times and longer breathing periods were associated with survival, possibly because patients with smaller tumours and better performance status tend to breathe slower. This finding is opposite to Sentker, where longer periods caused under-sampling artifacts, possibly because of differences in scan protocol. In multivariable analysis the effect disappeared except for a negative effect of long scan times that may be related to missing information when a full scan cannot be completed. Manufacturers should fix this.