Model-based cardiac dose estimation in radiation treatment of left breast cancer

Correlation between maximum heart distance and thoracic diameter changes and diaphragmatic descent in left-sided breast cancer patients during deep inspiration breath-hold (DIBH)

BACKGROUND

Cardioprotection is valued in radiotherapy for patients with left-sided breast cancer. Deep inspiration breath-hold (DIBH) technique can achieve cardioprotection well. However, during DIBH, the extent to which the heart enters the radiation field is affected by the movement of the thorax and diaphragm. The aim of this study was to analyze the correlation between the maximum distance of the heart entering the field (maximum heart distance, MHD) and thoracic diameter changes and diaphragmatic descent in left-sided breast cancer patients during DIBH.

PATIENTS AND METHODS

Ninety-eight patients with left-sided breast cancer were included in this retrospective study. They performed simulation in Sentinel-guided DIBH, and two sets of CT images were collected under both free breathing (FB) and DIBH, and diaphragm positions, anteroposterior thoracic diameter (ATD), transverse thoracic diameter (TTD), gating window level (GWL), and MHD were measured, and the change (Δ) of each parameter in DIBH relative to that in FB were calculated. Pearson or Spearman test were used to analyze the correlation between ΔMHD and the changes in other parameters.

RESULTS

For all patients with DIBH, the average of ΔMHD was -8.3 mm, and the average of ΔATD and ΔTTD were 11.0 and 8.6 mm, and the median of both left diaphragmatic descent (LDD) and right diaphragmatic descent (RDD) were 35.0 mm, and the median of GWL was 11.1 mm. The correlation coefficients between MHD decrease (ΔMHD) and LDD, RDD, and ΔTTD were -0.430 (p = 0.000), -0.592 (p = 0.000) and 0.208 (p = 0.040), respectively, but not significantly correlated with ΔATD or GWL.

CONCLUSIONS

The MHD decrease showed a moderate correlation with diaphragmatic descent In Sentinel-guided DIBH for patients with left-sided breast cancer, while there was a weak or no correlation with thoracic diameter changes or GWL. Abdominal breathing can lower diaphragm more and may be more beneficial to the heart stay away from tangential field.

Development of deep learning chest X-ray model for cardiac dose prediction in left-sided breast cancer radiotherapy

Deep inspiration breath-hold (DIBH) is widely used to reduce the cardiac dose in left-sided breast cancer radiotherapy. This study aimed to develop a deep learning chest X-ray model for cardiac dose prediction to select patients with a potentially high risk of cardiac irradiation and need for DIBH radiotherapy. We used 103 pairs of anteroposterior and lateral chest X-ray data of left-sided breast cancer patients (training cohort: n = 59, validation cohort: n = 19, test cohort: n = 25). All patients underwent breast-conserving surgery followed by DIBH radiotherapy: the treatment plan consisted of three-dimensional, two opposing tangential radiation fields. The prescription dose of the planning target volume was 42.56 Gy in 16 fractions. A convolutional neural network-based regression model was developed to predict the mean heart dose (∆MHD) reduction between free-breathing (MHD_FB) and DIBH. The model performance is evaluated as a binary classifier by setting the cutoff value of ∆MHD > 1 Gy. The patient characteristics were as follows: the median (IQR) age was 52 (47–61) years, MHD_FB was 1.75 (1.14–2.47) Gy, and ∆MHD was 1.00 (0.52–1.64) Gy. The classification performance of the developed model showed a sensitivity of 85.7%, specificity of 90.9%, a positive predictive value of 92.3%, a negative predictive value of 83.3%, and a diagnostic accuracy of 88.0%. The AUC value of the ROC curve was 0.864. The proposed model could predict ∆MHD in breast radiotherapy, suggesting the potential of a classifier in which patients are more desirable for DIBH.

Preoperative spirometry and BMI in deep inspiration breath-hold radiotherapy: the early detection of cardiac and lung dose predictors without radiation exposure

Background

This study aimed to investigate preoperative spirometry and BMI as early predictors of the mean heart and lung dose (MHD, MLD) in deep inspiration breath-hold (DIBH) radiotherapy.

Methods

Left-sided breast cancer patients underwent breast-conserving surgery followed by DIBH radiotherapy enrolled. Patients who were not available for preoperative spirometry were excluded. One hundred eligible patients were performed free-breathing (FB-) CT and DIBH-CT for plan comparison. We completed the correlative and multivariate analysis to develop the linear regression models for dose prediction. The residuals were calculated to explore the unpreferable subgroups and compare the prediction accuracy.

Results

Among the parameters, vital capacity (VC) and BMI showed the strongest negative correlation with MHD (r = − 0.33) and MLD (r = − 0.34), respectively. They were also significant in multivariate analysis (P < 0.001). Elderly and less VC were independent predictors of increasing absolute residuals (AR). The VC model showed no significant difference in AR compared to the model using the CT parameter of lung volume in FB (LV-FB): median AR of the LV-FB model vs. the VC model was 0.12 vs. 0.11 Gy (P = 0.79). On the other hand, the median AR of the MLD model was 0.38 Gy, showing no specific subgroups of larger AR.

Conclusion

Preoperative spirometry and BMI are significant predictors of MHD and MLD, respectively. Although elderly and low-VC patients may have larger predictive variations, spirometry might be a substitute for LV-FB as a predictor of MHD.

Dosimetric comparison between different tangential field arrangements during left-sided breast cancer radiotherapy

This study aimed to evaluate variations in dose distribution within the target volume and dose received by the organs at risk (OARs) for different tangential field arrangements during three-dimensional (3D) conformal treatment planning for left-sided breast cancer. Computed tomography (CT) images of 25 breast cancer patients were included, and three different mono-isocentric half-block (MIHB) treatment plans-parallel central axis technique (PCAXT), posterior border parallel technique (PBPT), and parallel quadrant technique (PQUDT)-were considered for each patient. The dosimetric and geometric parameters related to each followed plan were then extracted for the planning target volume (PTV) and the OARs, and compared. The results showed no significant differences among the extracted dosimetric and geometric parameters of the OARs for the different plans, while the D_max, V_95%, homogeneity index (HI), and conformity index (CI) values related to the PTV were significantly different (P < 0.05). The lowest D_max and V_95% values inside the PTV were related to the PCAXT plan. The best HI was achieved with the PBPT plan, whereas the best CI was observed for the PCAXT plan. The best correlation between the geometric and dosimetric parameters of the OARs was between V_5Gy-central lung distance for the ipsilateral lung and the V_5Gy-maximum heart distance for the heart in all plans. These results demonstrate that variations in the tangential field arrangement at the posterior border for optimal coverage of the PTV may not considerably affect the dose received by the OARs.

Fully automated detection of heart irradiation in cine MV images acquired during breast cancer radiotherapy

PURPOSE

To develop robust automated detection of heart irradiation in continuous portal images (cine MV images) of tangential breast cancer treatments.

METHODS

Cine MV images of 302 tangential field deliveries were recorded for ten left-sided breast cancer patients receiving deep-inspiration breath-hold radiotherapy. An algorithm for fully automated heart edge detection in cine MV images was developed and tested for all images. The algorithm first enhances the heart edge contrast greatly by exploiting that pixels on the heart edge change their intensity cyclically, and highly correlated, at 1-3 Hz due to heartbeat. The algorithm then detects the heart edge in the enhanced image and calculates the exposed heart area within the field aperture.

RESULTS

The algorithm correctly identified the heart edge in all cine MV series with heart exposure (169 of 302 field deliveries). With conservative selection criteria the algorithm on average identified 70 heart edge pixels in the heart-including field deliveries (range: 10-230) without false positives. With less strict criteria 106 heart edge pixels were identified on average (range: 13-262) with 0.6% being false positives. The heart edge bordering the lung was segmented highly reliably even a few millimeters outside the field edge. For six patients with frequent heart irradiation, the exposed heart area showed large interfraction variations and smaller intrafraction variations.

CONCLUSIONS

Automated heart edge detection in cine MV images was proposed, developed and shown to be highly efficient for heart exposure detection in tangential breast fields. It may allow unsupervised surveillance of heart exposure at all tangential breast cancer treatments in a clinic.