Evaluate the Feasibility of Gd-based Contrast Clearance Differenceto Delineate Subvolume Target in Primary and Metastatic Brain Tumors Radiotherapy- A Prospective Study


 Background: To evaluate the feasibility ofdelineating subvolume target in brain tumor radiotherapy using gd-based contrast clearance difference.Methods：Twenty-six patients with malignant brain tumor were scanned with MRI. The first and second acquisitions of standard T2-weighted images (T2WI) andT1-weighted images (T1WI) were respectively performed?> at 5 minutes and 60 minutesafter injection of contrast agent. Delayed contrast extravasation MRI(DCEM) computed by Brainlab concludesregions of contrast agent clearance which represent active tumor，andregions of contrast accumulation which represent non-tumor tissues. Based on T2WI images,14 patients were divided into group A and group B, with andwithout liquefaction necrosis, respectively. Then,gross target volume (GTV) was delineated on T1WI images. Based on the GTV, active tumor (GTV tumor) and non-tumorregions(GTV non-tumor) were delineated on T1WI-DCEM fusion images, whileliquefaction necrosis (GTVliquefaction)and non-liquefaction(GTVnon-liquefaction)were delineated on T1-T2WI fusion images. Finally, the differences between different subvolumes were compared by paired t-test.Results：In group A,the mean value of GTVA was 21.38±25.70 cm3, and the GTVnon-liquefaction and GTVliquefaction were 13.65±18.15cm3 and6.30±7.57cm3, respectively. The GTV tumor was 10.40±13.52 cm3 whilethe GTV non-tumor was 9.55±14.57 cm3, The GTVnon-liquefaction increased by an average of 28.2%（P＜0.05）,compared to GTV tumor . While the GTV non-tumor increased by an average of 46.3% (P＜0.05), compared tothe GTVliquefaction.In group B, the mean value of GTVB on enhanced T1WI was4.39±3.75 cm3. The GTV non-tumorreduced by an average of 50.3% (P＜0.05) , compared totheGTV tumor.Conclusion：Comparedto T2WI, the DCEM has advantages in identifyingthe liquefaction areaand could clearly differentiatesubvolume of active tumor from non-liquefaction necrosis.DCEMis meaningful in guiding the delineation of subvolume in primary and metastatic brain tumors.


Background
Radiotherapy is one of the indispensable treatment methods for brain tumor, whose effect is one of the signi cant factors affecting the survival of patients. With the development of radiotherapy technology, radiotherapy for brain tumors has achieved encouraging results. Magnetic resonance has become a routine method in the radiotherapy management of brain tumor patients by virtue of its high resolution of soft tissue [1][2][3]. At present, it is crucial to accurately identify tumor tissues from non-tumor tissues for individualized treatment of brain tumor patients, but traditional MR cannot distinguish the effects of tumor progression and treatment induction. Multimodal magnetic resonance imaging, such as MRS PWI,could provide clinician with information which concluded tumor metabolism, pathophysiology, microcirculation state [4][5]. Combining the multimodal magnetic resonance imaging made the application of MR technique become more and more widely in detecting focus, guiding the tumor target outline and observing therapeutic response dynamically. However, radiotherapy failure still occured from time to time, which may be related to the different sensitivity in multimodal magnetic resonance imaging and insu cient local Radiotherapy dose of tumors. In the era of precise radiotherapy, it was a research hotspot that identifying the biological characteristics of tumors in different regions before radiotherapy to achieve the intensity modulated radiotherapy within the tumor. PET was relatively insensitive to brain tumors, expensive and radioactive, which limited its wide application. so we may achieve biological IMRT(BIMRT) by applying MR to observe brain tumor characteristics. Leor et al. [6] showed that gd-based contrast clearance difference could provide reliable tumor load information and own potential value in predicting brain tumor recurrence. Therefore, we carried out a study that applyed gd-based contrast clearance difference on to delineate subvolume brain tumor and provided a basis for the design of future radiotherapy plans.

Patients
The study include 26 patients (mean age: 57 ± 12 years old) with brain tumor, 25 cases of patients with metastatic tumors, a case of patient with primary tumor, 14 men and 12 women, 37 tumor lesions. None of the patients received surgical resection in tumor area. Every tumor exceeded 0.5 cm 3 .The study was approved by the Institutional Ethics Committee of Shandong Cancer Hospital and Institute, and written informed consent was obtained from all patients.

Mri Experiments
All MRI experiments were performed at 3T MR system (Discovery 750w, GE Healthcare, USA) with brain coil. For the T1WI measurement, the scan parameters were applied as follows: TR = 8.47 ms, TE = 3.25 ms, matrix = 256 × 256, FOV = 256 mm x256mm, slice thickness = 1 mm. For the T2WI measurement, the scan parameters were applied as follows: TR = 13312 ms, TE = 113.5/EF, FOV:26 × 26, slice thickness = 3 mm. A standard single dose (0.1 mmol/kg) of gadolinium DOTA was injected intravenously. MR sequences included T2 PROPELLER, T2 FLAIR and echo-planar diffusion-weighted MRI (DWI). The rst acquisitions of standard T1-weighted images (T1WI) and T2-weighted images (T2WI) were performed at 5 minutes after injection of contrast agent while the second acquisitions of T1WI were performed at 60 minutes after injection of contrast agent.

DCEM acquisition
All images were uploaded to Brainlab software whose contrast clearance analysis module was performed by subtracting T1WI acquired 60 minutes after Gadolinium(Gd)-based contrast agent injection from those

Image processing
Based on different signal changes on T2WI images, patients were divided into group A which own T2WI high signal and group B, with and without liquefaction necrosis, respectively. DCEM were registrated rigidiy with the rst T1WI enhanced image and T2WI in fusion module of Brainlab. In order to achieve accurate registration, the researchers can carry out manual adjustment or ne-tuning.
Then, gross target volume (GTV) was delineated on rst T1WI images. Based on the GTV, active tumor (GTV tumor ) and non-tumor regions (GTV non−tumor ) were delineated on T1WI-DCEM fusion images, while liquefaction necrosis (GTV liquefaction ) and non-liquefaction (GTV non−liquefaction ) were delineated on T1-T2WI fusion images. All the target were delineated by the same researcher in Smartbrush module of Brainlab software. The volume of each target area was automatically generated by the software. Finally, the volume and statistical differences of different subvolume were calculated (Fig. 1).

Statistics
Data were analyzed using SPSS25.0 and represented mean ± standard deviation. Paired t-tests was used to determine the difference of sub-volume between DCEM and T2WI. Signi cance threshold was set as p < 0.05.

Group informations
The mean value of GTV total with 26 patients was (10.28 ± 17.25) cm 3

Subvolume Target Informations
In group A, the non-liquefaction and liquefaction necrosis on T2WI were 13.65 ± 18.15 cm 3 and6.30 ± 7.57 cm 3 , respectively. Active tumor area was 10.40 ± 13.52 cm 3 while the non-tumor area was 9.55 ± 14.57 cm 3 , The non-liquefaction necrosis volume increased by an average of 28.2%(P 0.05 ; Table 3 and Fig. 2), compared to active tumor area. It suggested that the solid component region of T2WI was not completely the region with tumor activity. While the non-tumor tissues increased by an average of 46.3% (P 0.05), compared to the liquefactive necrosis tissues. It showed that DCEM could nd the potential area of liquefactive necrosis which T2WI couldn't display (Fig. 3).
In group B, the mean value of GTV B on enhanced T1WI was 4.39 ± 3.75 cm 3 . The GTV non−tumor reduced by an average of 50.3% (P 0.05 Table 4), compared to the GTV tumor . It found that The detection of small brain tumors by DCEM still revealed tiny potential areas of liquefactive necrosis which T2WI couldn't detect.

Discussion
Cerebral metastases and glioma were the most common and highest aggressive malignant tumor, that could be locally controlled by radiation therapy, so as to slow down the disease progression and prolong survival time of patients. Although intensity-modulated radiation therapy has reached uniform distribution of dose in tumor area, but Radiotherapy failure still exists, on account of homogeneous distribution of tumor cells [7] and differences in radiation sensitivity.
Hazle et al. [8] found that radioactive necrosis and tumor tissues were enhanced at different rates by scanning T1WI enhanced images after a delay of 10-15 minutes, and further identi ed the recurrent tumor and radioactive necrosis area by delayed scanning. Leor et al. [9] suggested that the longer you delayed, the easier you will nd tumor tissue, respectively, comparing T1WI enhanced images of 30 patients with the delay of 15 minutes and 75 minutes with postoperative samples. There was no signi cant difference between primary and metastatic brain tumors. One or more of the following characteristics of tumor morphology were the tumor tissue activity criteria in DCEM: cellularity, small cells, mitosis, high Ki67, pseudo-palisading necrosis and vascular proliferation. The criteria for non-tumor activity were the following: large, widely spaced atypical astrocytes, blood vessels hyalinization, brinoid material in vessels, proliferating small vessels and non-palisading tumor necrosis. Since each patient needed two T1WI enhanced scanning, In order to avoid the discomfort of waiting for a long time on the MR machine, we asked the patients to leave the machine after the rst MR scan and perform the second scan after resting for 1 hour, so as to improve the e ciency of the experiment. The operator positioning errors could reach more than 5 , based on the subjective experience and xed thermoplastic mask, while the allowable error of the positioning xator is generally ± 2 mm in radiotherapy. In order to achieve su ciently accurate degree in the experiment, we recruited 26 patients who have undergone CT simulation positioning. Same operators could reduce the involuntary movement and increase scanning repeatability by mask production that could be used to mark relative position between the mesh of mask and skin of forehead and quality control of every aspect.
Evaluation means of radiotherapy failure were relatively limited at present. Conventional MR has apparently not distinguished active area and high risk of recurrence from tumor area, especially In the case of complex situation that conclude broplasia and peritumoral edema. Weybright et al. [10,11] found that Cho/Cr > 2 and (or) Cho/NAA > 2. 5 prompted tumor recurrence by means of studying metabolites of recurrent gliomas, but whose disadvantage was that not independent indicator for diagnosis. Functional MRI which concluded DSC-MRI and DCE-MRI could effectively evaluate vascular permeability and angiogenesis [12]. Some studies have shown that value of relative cerebral blood volume (rCBV) could be seen as one of the most valuable parameters for evaluating tumor classi cation [13]. Wang et al. [14] found that glioma had no recurrence according to rCBV < 1 by means of perfusion weighted imaging (PWI), while rCBV > 2 suggested greater possibility of recurrence, DSE based on PWI could assess tumor status and predict tumor behavior. ASL PWI based could predict blood vessel density, which was strongly correlated with tumor and more stable than DCEM [15,16]. Leor et al. [6,9] found that the regions where contrast agent clear rapidly in DCEM were signi cantly correlated with high rCBV.
However, enhanced lesions with high rCBV had no statistical difference comparing with regions of interest (ROI) which were the same shape and size on the opposite normal brain organization. It suggested that high CBV in opposite normal brain anatomy may hide the accuracy of functional magnetic resonance. The high rCBV in the tumor region was statistically different from the contralateral ROI when tumor location separated from gray matter. The mean blue volume of the enhanced tumor region on DCEM was statistically signi cantly different from contralateral ROI. It indicated that DCEM was less affected by the normal contralateral tissue. Leor et al. also found that DCEM, whose sensitivity was up to 90% in detecting active tumors, could clearly show the small lesions which were not detected by DSC-MRI and couldn't nd artifacts which were prone to appear due to anatomical factors in DSC-MRI.
This study, based on distinguishing tumor activity from DCEM and liquefactive necrosis from T2WI, found that DCEM could nd the potential areas of liquefactive necrosis which T2WI couldn't display.
Tumor cells can't proliferate owe to reduction of blood supply which was due to Immature vessel, Clogged vascular lumen by endothelial cell and medical intervention such as radiotherapy, chemotherapy and so on. It bought about which tumor cells with earlier blood disruption slowly evolved into liquefaction necrosis. Liquefaction necrosis was high signal in T2WI, on account of high sensitivity to free water. However, tumor cells which were neither activation nor blood supply which disappeared later didn't become liquefaction. It made some part of GTV non−tumor be disappeared in T2WI. Based on this conclusion, DCEM could provide the trend or degree of tumor necrosis for clinicians. Of course, tumor cells without blood supply may become brosis. A patient with brain metastases from breast cancer existed in the phenomenon where GTV liquefaction (13 cm 3 )were larger than GTV non−tumor 10.1 cm 3 in group A. We thought that it is large volume of free water generated by large necrotic area that covered the part of active tumor area, so as to show the larger necrotic in T2WI.
According to the statistical data of group B, the average volume of tumor was smaller than group A. Although T2WI did not indicate liquefaction necrosis, but the GTV non−tumor could be detected on DCEM, on the one hand it further proved the advantage in nding potential necrosis area, on the other hand this advantage was also applicable to small brain tumors. This conclusion was similar to Leor's opinion that DCEM is more sensitive to MR in detecting small brain tumors. In group A, GTV non−liquefaction was larger than GTV tumor , in other words, the active tumor area was not completely the solid tumor component on T2WI. Leor's study showed that the clearance differences of Gd-based contrast agents could re ect reliable tumor load information and had potential value in predicting brain tumor recurrence. Therefore, we considered that GTV tumor had higher risk of recurrence than other solid tumor regions. It was hysteretic to predict the prone relapse areas of tumor in DCEM. It was more signi cant to control recurrence areas of high risk in advance. Therefore, we were able to delineate GTV tumor before radiotherapy and to design a radiotherapy plan with a higher dose so as to improve local control of the tumor. Because according to Walker's report [17,18], there was an obvious dose-response relationship that the survival rate of patients would improve with the increase of radiotherapy dose in tumor areas.
In clinical application, there was no doubt that it still existed in many problems, such as the improvement of image registration accuracy, the choice of dose in the high-risk area, the choice of tumor type and the progression tendency of non-tumor areas which were beyond the necrotic areas. In addition, In order to apply DCEM to clinical radiotherapy, we were supposed to evaluate it in all directions, combining with multidisciplinary and multi-image.

Conclusion
Compared to T2WI, the DCEM has advantages in identifying the liquefaction area and could clearly differentiate subvolume of active tumor from non-liquefaction necrosis. DCEM is meaningful in guiding the delineation of subvolume in primary and metastatic brain tumors. Availability of data and materials

Abbreviations
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions
YXWand GZG contributed to the study design, the patient enrollment, the data statistics and analysis and writing the manuscript.JL and YY contributed to reviewing the delineation. YS and LZW made important contributions in collecting the data and revising the content. All authors read and approved the nal manuscript.
Ethics approval and consent to participate  Supplementary Files This is a list of supplementary les associated with this preprint. Click to download. Coverletter.docx Supportinformation.docx