Preliminary Study on the Effect of 4DCT-Ventilation-Weighted Dose on the Radiation Induced Pneumonia Probability (RIPP).

PURPOSE: The purpose of the present study was to evaluate the feasibility of using 4-dimensional computed tomography (4DCT)-ventilation-weighted dose analysis to predict radiation-induced pneumonia probability (RIPP). METHODS AND MATERIALS: The study population for this retrospective analysis included 16 patients with stage III lung cancer. Each patient's 4DCT images, including end-inhale and end-exhale sequences, were used for the deformable image registration, and the Hounsfield units (HU) density-change was used to calculate the ventilation. A previously established equation was used to convert the original dose (OD) D 0, i in the lungs in the original plan (OP) to the weighted-dose (WD) D w, i in the weighted plan (WP). The patients were divided into 2 groups, one with radiation-induced pneumonia (RIP), and one without. The Spearman correlation analysis was used to analyze the correlation of RIP with ΔV20 (ΔV x = V w, x in the WP - V 0, x in the OP), ΔMLD (ΔMLD = mean lung dose (MLD) in the WP - MLD in the OP), and ΔV5.
RESULTS: The results showed that 5 of the 16 patients were suffering from acute RIP, 4 of which had higher ΔV20 and ΔMLD values than the rest of the patients. The results of the Spearman correlation analysis for those 4 patients were as follows: RIP vs. ΔV20, r = 0.5123; RIP vs. ΔMLD, r = 0.5119; RIP vs. ΔV5, r = 0.1904.
CONCLUSIONS: The 4DCT-ventilation-based weighted-dose analysis showed some correlation between RIPP and both ΔV20 and ΔMLD, when comparing the weighted-dose and the conventional dose-volume histogram (DVH) analyses.

Introduction

Radiotherapy plays an important role in the comprehensive treatment of lung cancer. Hypofractionated radiotherapy is the best treatment for stage I-II (T1-3N0-T1-2N1) lung cancer patients, especially for elderly patients with comorbidities who are not ideal candidates for surgery. Previous studies have shown that stereotactic body radiation therapy (SBRT) results in fewer complications, but has almost the same local control rate as surgery. Concurrent chemoradiotherapy is recommended for locally advanced stage II-III lung cancer, while radiotherapy is considered a local consolidation therapy for stage IV lung cancer metastases.

The tolerance dose of lung tissue is still the primary constraint for improving the target dose. The radiation dose to the lung tissue has continuously decreased with the growth and development of treatments and technology. One report has demonstrated that symptomatic radiation-induced pneumonia (RIP) and fibrosis were serious complications occurring in an estimated 5-50% of patients with lung cancer. The effect of the continuous improvement of radiotherapy techniques to further reduce the dose to the lung is decreasing. In this regard, functional lung avoidance radiotherapy has been proposed. In this process, the lung is first divided into a high functional area (HFA) and a low functional area (LFA), based on functional imaging results. Then, the radiotherapy planning is designed to spare as much of the HFA as possible, based on the subarea, as one opinion suggested that sparing the HFA may reduce the occurrence of pulmonary complications.

Functional avoidance based on positron emission tomography (PET), single-photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), and 4-dimensional computed tomography (4DCT)-ventilation imaging has been demonstrated in a variety of studies. Of these methods, 4DCT-ventilation imaging, which uses data to create lung function maps for functional avoidance planning, is the most ideal. Compared with other imaging modalities, 4DCT-ventilation imaging provides functional information from standard treatment procedures without additional image acquisition, which does not increase the economic burden on the patient. Furthermore, 4DCT-ventilation imaging allows for rapid image processing without the need for a radioactive contrast agent, and provides a 4-D map of anatomical structures.

The history of development of 4DCT-ventilation for functional lung avoidance radiotherapy can be found in retrospective studies by institutions worldwide. However, previous studies involving 4DCT-ventilation imaging for functional lung avoidance have primarily focused on demonstrating the feasibility of sparing HFAs. Little information in current literature addresses the evaluation of dose-volume in relation to functional weight, because planning the avoidance of HFAs would inevitably increase the dose to LFAs. The acquisition of data regarding functional avoidance and loss of functional defect are expected to be obtained. The first goal of data acquisition is to evaluate the ratio of the functional area to the functional defect area. Since the HFA has a higher ventilation capacity and oxygen concentration than the LFA, and oxygen is known to increase tissue sensitization to radiation therapy, the Hounsfield unit (HU) changes, which indicate ventilation ability, are used as a weight for dose analysis. Therefore, we analyzed patients who had completed radiotherapy at the radiation oncology center of our hospital using 4DCT-ventilation imaging to determine the weight for a weighted-dose in order to explore the feasibility of the weighted-dose analysis for predicting RIP.

Materials and Methods

A total of 16 patients were included for analysis in the present study. The patient population consisted of 12 men and 4 women, aged between 46 and 66 years, with a median age of 56. Patients with squamous cell carcinoma (n = 6), adenocarcinoma (n = 6), and small-cell lung cancer (n = 4) received platinum-based double-drug chemotherapy with radiotherapy.

A 24-slice CT scanner (Siemens, Munich, Germany) was used to acquire 4-D data using an abdominal pressure respiratory induction system. The patients were placed in a supine position with their hands above their head, and were held in place by a thermoplastic reticular membrane during scanning. The respiratory cycle was divided into 10 phases for data collection. A CT sequence was imaged with the patient breathing freely. The anatomy was scanned from the clavicle to the liver. Each patient’s CT scan was reconstructed with a 3.0 mm slice thickness and transferred to the Pinnacle TPS 9.10.

All 16 patients were treated with a Trilogy linear accelerator (Varian Medical Systems, CA, USA) using a 6 MV photon beam. Intensity-modulated radiotherapy (IMRT) plans (named original plans, OPs) with 5-6 fields were designed for every patient, based on direct machine parameter optimization.

An OPs was designed for each of the 16 patients by the radiation oncology physicists once the internal target volume (ITV), planning target volume (PTV), and organs at risk (OAR) were defined by the doctors. The initial objective functions for the OPs are as follows: PTV – 60-66 Gy/30F, D95% > 60-66 Gy; heart – V30 < 40%, V40 < 30%; right lung and left lung – V5 < 65%, V20 < 30%; esophagus – V50 < 30%, V60 < 10%; spinal cord – maximum dose (MaxD) < 45 Gy. These initial objective functions were used during the optimization, during which the physicists modified the objective functions several times until individualized optimal objective functions for each patient were established. The OPs were designed using free-breathing CT imaging because similar dosimetric characteristics between the free-breathing and the average intensity projection CTs have been reported in previous literature. Moreover, we had been using the free-breathing CT as the basis of planning, and were quite sure that it was better for the OP design. The left and right lungs were then merged into the A-Lung for evaluation. (Note: D95% – minimum absorbed dose which reaches 95% of the target volume; V is the ratio between the volume of organs with absorbed a dose greater than x Gy and the total volume of organs)

A deformable image registration algorithm in MIM,(a software for imaging analysis, was used to register end-inhale CT sequences with the corresponding end-exhale CT sequences, in order to calculate a 4DCT-ventilation weighted-dose map for each patient.

In the present study, the end-inhale and end-exhale CT sequences were selected for the calculation of 4DCT-ventilation with a grid of 4.0 mm. The registration was used to calculate ventilation based on the HU density change, as follows :

where V and V are the end-inhale and end-exhale volumes, and HU and HU are the end-inhale and end-exhale Hounsfield units of the individual lung voxels, respectively. HUs were defined as the average value in the 4.0 mm-grid, as this size grid is sufficient for the dose calculation of conventional fraction radiotherapy. During CT image reconstruction for planning, with a slice thickness of 3.0 mm, the grid would be slightly larger than 3.0 mm when calculating the average HU and 4DCT-ventilation.

Assuming that the original dose (OD) in the OP was D and the weighted dose (WD) in the WP was D , the equation relating D and D was as follows:

which considers the change in lung function identified by ventilation in each grid to calculate the dose.

The end-inhale and the end-exhale CT sequence images were subtracted to achieve after registration, which created a ventilation map (VP) in A-Lung, as shown in Figure 1. Then, the VP and free-breathing CT underwent deformable image registration. The one-to-one correspondence between D and was used to calculate D , as shown in Figure 2.

Figure 1.

The end-inhale computed tomography (CT) sequence was set as the “Planning CT (fixed image)”, shown in the first layer. The end-exhale CT sequence was set as the “Diagnostic CT (floating image)”, shown in the second layer. The Hounsfield unit (HU) density-change map was achieved using deformable image registration and was subtracted, as shown in the third layer.

Figure 2.

Deformable image registration of the free-breathing CT, which was the original plan (OP) design, and the HU density change map. The established one-to-one correspondence between D and was used for calculating D

The occurrence of radiation-induced pneumonia is related to many factors, and it is of great clinical significance to control these factors. Studies have shown that the dose factors heavily associated with radiation-induced pneumonia are V , V , where ΔV V in the WP – V in the OP, and mean lung dose (MLD) calculated from A-Lung. Therefore, the parameters evaluated in the present study include V , V , and MLD. The corresponding values after weighted-dose calculation are V , V , and MLD , and their differences are ΔV (V -V ), ΔV (V -V ), and ΔMLD (MLD -MLD). The correlations between the differences, ΔV , ΔV , and ΔMLD, and the occurrence of radiation-induced pneumonia were analyzed.

Results

The average A-Lung values for V5, V20, and MLD of the 16 patients were 67.9 ± 7.5% (55.4-80.2%), 29.9 ± 2.9% (26.3-38.6%), and 1741 ± 152 cGy (1504-2172 cGy) in the Ops, versus 79.2 ± 6.8% (67.3-88.6%), 38.4 ± 2.7% (34.7-44.2%), and 1916 ± 147 cGy (1645-2290 cGy) in the WPs, respectively, as shown in Table 1.

Table 1.

Data Regarding the Original Dose (OD) in the Lung From the Original Plan (OP), the Weighted-Dose (WD) From the Weighted Plan (WP), and the Dose Difference (ΔD).a

Pat.	V5(%)	V20(%)	MLD (cGy)	V 5, w (%)	V 20, w (%)	MLD w (cGy)	ΔV 5 (%)	ΔV 20 (%)	ΔMLD (cGy)
1	80.2	30.2	1710	88.6	35.5	1854	8.4	5.3	144
2	59.3	31.4	1782	73.3	37.2	1931	14	5.8	149
3	55.4	26.6	1593	67.3	37.2	1889	11.9	10.6	296
4	78.6	38.6	2172	87.3	44.2	2290	8.7	5.6	118
5	62.2	28.2	1819	70.6	38.7	2025	8.4	10.5	206
6	73.1	31.9	1790	81.9	41.9	1911	8.8	10.0	121
7	68.5	27.7	1691	84.5	35.6	1843	16.0	7.9	152
8	65.7	28.5	1630	78.7	40.2	1863	13.0	11.7	233
9	74.3	32.2	1942	87.4	39.4	2132	13.1	7.2	190
10	72.4	31.3	1785	83.1	39.6	1953	10.7	8.3	168
11	68.5	30.0	1760	81.4	42.6	2018	12.9	12.6	258
12	67.2	26.3	1620	78.8	34.7	1751	11.6	8.4	131
13	77.5	30.9	1776	86.4	37.1	1886	8.9	6.2	110
14	65.8	30.2	1504	76.3	36.4	1645	10.5	6.2	141
15	56.5	27.7	1622	69.5	39.1	1890	13.0	11.4	268
16	61.5	27.3	1655	71.9	35.4	1778	10.4	8.1	123
Mean	67.9	29.9	1741	79.2	38.4	1916
s	7.5	2.9	152	6.8	2.7	147

aΔV (V -V ), ΔV (V -V ), ΔMLD (MLD -MLD).

Based on the clinical symptoms of adverse events (RIP) described in the Common Terminology Criteria for Adverse Events 4.0 (CTCAE4.0), 3 radiation oncologists on the same team assessed and graded RIP. As shown in Table 2, 5 of the 16 patients had acute RIP (pneumonia which occurred < 1 month after the completion of radiotherapy), 4 of which (patients 3, 8, 11, and 15) had grade III RIP (patient required oxygen) and higher ΔV5, ΔV20, and ΔMLD values compared to those of the other 12 patients, as shown in Table 1. However, the V5, V20, and MLD values for those 4 patients were not at the top of the OPs. The Spearman correlation analysis was used to analyze the correlation between RIP and ΔV20, ΔMLD, and ΔV5. The results were as follows: RIP vs. ΔV20, r = 0.5123; RIP vs. ΔMLD, r = 0.5119; and RIP vs. ΔV5, r = 0.1904.

Table 2.

Pneumonia Statistics Table.a

Patient	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
Pneumonia			√	√				√			√				√

a “√” lables patients suffering from pneumonia.

Discussion

RIP is one of the major limitations of radiotherapy for patients with lung cancer. It is irreversible and may interrupt treatment, while at the same time, serious pneumonia may endanger the patient’s life. RIP is influenced by a variety of factors. As of yet, the mechanisms of RIP are not well understood. The widely-accepted view at this time is that RIP is caused by damaged type II epithelial and endothelial cells and a series of signals from numerous acute inflammatory cells in locally damaged lungs. Irrespective of the pneumonia pathogenesis, one important clinical phenomenon is the decline in ventilatory function once radiation-induced pneumonia occurs.

As functional imaging was lacking in the past, radiation oncologists held the opinion that the lung was a typical parallel organ, and that the function of each part was homogeneous, as was the radiation sensitivity. Thus, dose assessment usually focuses on the V and MLD of the entire lung, although dose location has not yet garnered much attention. Although there have been studies conducted based on the division of the lung into upper and lower parts, these results were rarely used in clinical practice.

After the development of functional imaging, lung function could be clearly evaluated, and the lung could be divided into the HFA and LFA, corresponding to good and poor ventilation, respectively. The information regarding HFAs and LFAs provided by lung-function imaging was used to make the therapeutic dose as small as possible in the HFA and protect the HFA in the design of the radiotherapy plan. The occurrence rate and severity of RIP are expected to be reduced by this method. However, one thing that was not demonstrated was the acquisition of HFA avoidance rather than loss of LFA involvement. This is very important, because no matter how the plans were designed, it was impossible to avoid HFAs completely, especially in cases with tumors located in HFAs. It is of note, however, that even in the HFA and LFA, lung function was not homogenous.

In the present study, we did not set any certain parameter by which to distinguish lung tissue into HFA and LFA, based on this knowledge. The function of each voxel/grid was labeled with 4DCT-ventilation, in an attempt to have the “label” reflect the dose of the voxel/grid. The results indicated that this method might be useful. Of the 16 patients, 5 patients (patients 3, 4, 8, 11, and 15) suffered from RIP. We considered V5, V20, and MLD to be closely associated with RIP, and found that the V5 of the OPs in patients 3, 4, 8, 11, and 15 ranked (from high to low) 16th, 2nd, 10th, 8th, and 15th, of 16 patients, V20 ranked 15th, 1st, 10th, 9th, and 13th, and MLD ranked 15th, 1st, 14th, 8th, and 13th. In other words, these patients did not rank in the top in for V5, V20, or MLD; however, they suffered from RIP, while the other 11 patients with higher V5, V20, and MLD values did not.

When V , V , and MLD were evaluated, it was easy to find that the ranking of ΔV , ΔV , and ΔMLD for patients 3, 4, 8, 11, and 15 were at the top, particularly the ΔV and ΔMLD for patients 3, 8, 11, and 15. For patients 3, 8, 11, and 15, the rate of high-dose radiation exposure to the HFA was higher than that of the other patients. Subsequently, the HFA of the lung was more seriously damaged, which undoubtedly increased the probability of developing RIP. The results of the Spearman correlation analysis corroborated this, as for RIP vs. ΔV20, r = 0.5123; RIP vs. ΔMLD, r = 0 .5119; and RIP vs. ΔV5, r = 0.1904. Spearman’s rank-order correlation is a method used to study the correlation between 2 variables according to ranking data. The RIP correlation coefficients for ΔV20 and ΔMLD were greater than 0.5, indicating a certain correlation between RIP and both ΔV20 and ΔMLD.

Of course, there were individual differences in the realized clinical probability. It was evident that of the 16 patients, the ΔV20 and ΔMLD for patient 5 were higher, however, the patient didn’t suffer from acute RIP.

There were many points worth discussing, if the weighted method was used to explain the clinical results of these 16 patients, such as how equation (2) expressed the weighted dose. By comparing equations (1) and (2), we found that equation (2) reflected the amount of air exchange in the grid (4.0 mm × 4.0 mm × 4.0 mm). The results of the present study showed that radiation sensitivity was positively correlated with oxygen amount, regardless of cells and tissues. The ΔV20 and ΔMLD values found in the present study were produced by equation (2). Namely, the ΔV20 and ΔMLD were clinical (macro) manifestations of “oxygen concentration-increasing-radiosensitivity.”

However, the number of patients in the present study was small, and further studies are needed to evaluate whether this approach would have the same results in a larger sample. Additionally, there may be other factors that could have induced errors in the analysis, such as the CT used in the studies, artifacts, and ventilation reconstruction algorithms, which depend on the improvement of image processing technology and algorithms.

Conclusion

The 4DCT-ventilation-based weighted-dose analysis in the present study showed that there was some correlation between RIPP and both ΔV20 and ΔMLD, when comparing the weighted-dose and conventional DVH analyses.