|Ahead of print publication
Magnetic resonance imaging radiomic feature analysis of radiation-induced femoral head changes in prostate cancer radiotherapy
Hamid Abdollahi1, Seied Rabi Mahdavi2, Isaac Shiri3, Bahram Mofid4, Mohsen Bakhshandeh5, Kazem Rahmani6
1 Department of Medical Physics, School of Medicine, Iran University of Medical Sciences, Tehran, Iran
2 Department of Medical Physics, School of Medicine, Iran University of Medical Sciences; Radiation Biology Research Center, Iran University of Medical Sciences, Tehran, Iran
3 Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences; Department of Biomedical and Health Informatics, Rajaei Cardiovascular, Medical and Research Center, Iran University of Medical Sciences, Tehran, Iran
4 Department of Clinical Oncology, Shohada-e-Tajrish Medical Center, Shahid Beheshti University of Medical Sciences, Tehran, Iran
5 Department of Radiation Technology, Allied Medicine Faculty, Shahid Beheshti University of Medical Sciences, Tehran, Iran
6 Department of Biostatistics, School of Public Health, Iran University of Medical Sciences, Tehran, Iran
Department of Medical Physics, School of Medicine, Iran University of Medical Sciences, Junction of Shahid Hemmat and Shahid Chamran Expressways, Tehran 14496
Seied Rabi Mahdavi,
Department of Medical Physics, School of Medicine, Iran University of Medical Sciences, Tehran
Source of Support: None, Conflict of Interest: None
Background and Purpose: As a feasible approach, radiotherapy has a great role in prostate cancer (Pca) management. However, Pca patients have an increased risk of femoral head damages including fractures after radiotherapy. The mechanisms of these complications are unknown and time of manifestations is too long; however, they may be predicted by early imaging. The main purpose of this study was to assess the early changes in femoral heads in Pca patients treated with intensity-modulated radiation therapy (IMRT) using multiparametric magnetic resonance imaging (mpMRI) radiomic feature analysis.
Materials and Methods: Thirty Pca patients treated with IMRT were included in the study. All patients underwent two mpMRI pre- and postradiotherapy. Thirty-four robust radiomic features were extracted from T1, T2, and apparent diffusion coefficient (ADC) obtained from diffusion-weighted images. Wilcoxon signed-rank test was performed to assess the significance of the change in the mean T1, T2, and ADC radiomic features postradiotherapy relative to preradiotherapy values. The percentage change values were normalized based on the natural logarithm base ten. Features were also ranked based on their median changes.
Results: Sixty femoral heads were analyzed. All radiomic features have undergone changes. Significant postradiotherapy radiomic feature changes were observed in 20 and 5 T1- and T2-weighted radiomic features, respectively (P < 0.05). ADC features did not vary significantly postradiotherapy. The mean radiation dose received by femoral heads was 40 Gy. No fractures were observed within the follow-up time. Different features were found as high ranked among T1, T2, and ADC images.
Conclusion: Early structural change analysis using radiomic features may contribute to predict postradiotherapy fracture in Pca patients. These features can be identified as being potentially important imaging biomarkers for predicting radiotherapy-induced femoral changes.
Keywords: Feature changes, femoral head, magnetic resonance imaging, prostate cancer, radiomics, radiotherapy
|How to cite this URL:|
Abdollahi H, Mahdavi SR, Shiri I, Mofid B, Bakhshandeh M, Rahmani K. Magnetic resonance imaging radiomic feature analysis of radiation-induced femoral head changes in prostate cancer radiotherapy. J Can Res Ther [Epub ahead of print] [cited 2019 Feb 19]. Available from: http://www.cancerjournal.net/preprintarticle.asp?id=251397
| > Introduction|| |
Prostate cancer (Pca) is the most common malignancy among men. As a feasible approach, radiotherapy has a great role in Pca management. However, normal tissue toxicities during or after radiotherapy remain the main limiting factors with negative impacts on patient's quality of life. Radiation-associated femoral head/neck toxicities including fractures, cortical bone thinning, and necrosis are complications which may occur months to years after radiotherapy., Previous studies have indicated that femoral insufficiency fracture is significantly higher in patients who receive pelvic irradiation., Femoral head fractures due to radiotherapy are associated with high morbidity and significant mortality.
Although the mechanisms by which radiation induces bone damages have been studied extensively, they remain unclear. Some studies have revealed that radiotherapy changes the biomechanical properties of bones including matrix embrittlement. Furthermore, previous research has shown that radiotherapy suppresses normal osteoblast proliferation. It was also shown that radiation sensitizes bone cells to apoptosis.
Magnetic resonance imaging (MRI) can be utilized as a feasible modality to detect bone abnormities., MR image sequences including T1-weighted (T1W), T2W, and diffusion-weighted (DW) are of importance in bone disease diagnosis. On the other hand, apparent diffusion coefficient (ADC), which is proportional to the cellular density, ratio of extracellular and intracellular components, and also soluble intracellular macromolecules, has been used in differentiating and detecting bone pathologies.
Recent advances in image analysis and biomarker research have produced powerful platforms for diagnosis, prognosis prediction, and therapeutic management of diseases. As an important current and future-based science, radiomics have been used to quantify tumor characteristics and also radiotherapy-associated clinical manifestations., It is a newly accepted branch of image processing which aims to correlate high-dimensional imaging features with clinical data. Furthermore, in radiomic literature, changes in different imaging features were correlated with radiation dose and clinical observations such as tissue shrinkage or density changes and hearing loss.,,
In the present study, for the first time, we applied radiomic feature analysis on MR images with the aim to assess changes in femoral head after the course of radiotherapy.
| > Material and Methods|| |
Our radiomic flowchart is depicted in [Figure 1]. From [Figure 1], our study is composed of four continuing phases including (1) imaging and segmentation, (2) image preprocessing, (3) feature extraction, and (4) data analysis.
In this prospective study, to test how radiotherapy induces femoral head changes, we conducted an MR image radiomic feature analysis. Between January 2015 and September 2016, we enrolled 30 Pca patients, with different Gleason scores, in accordance with the ethical standards of the responsible committee on human experimentation and in compliance with the 1975 Declaration of Helsinki and its revision in 2000. Because this work was conducted as prospective, all participants gave informed consent before enrollment. Patients were treated with intensity-modulated radiation therapy (IMRT), with 70.6 Gy in 25 fractions and treatment was delivered with multileaf collimators using 9 coplanar or noncoplanar, nonuniform beams. All planning was performed using Eclipse treatment planning system (Varian Medical Systems). In the plans, both right and left femoral heads were delineated as organs at risk and their mean radiation doses were obtained from TPS and their relevant dose volume histograms. The patients were followed for 6 months to 1 year for postradiotherapy bone damages. The patient details are shown in [Table 1].
Magnetic resonance imaging
All participants underwent two MRI scans, pre- and postradiotherapy. The mean time for MR scanning after radiotherapy was 3 months. Both pre- and postradiotherapy MRI scans included T1W, T2W, and DW imaging protocols. All studies were performed using 1.5 Tesla MRI machines (Siemens and Philips) equipped with endorectal and phased-array coils. The imaging protocols are shown in [Table 2]. For each patient, MR image protocols were the same in pre- and postradiotherapy scans. ADC was automatically calculated by the imager software with the use of the b-values reported in [Table 2].
Multiple radiomic feature sets including histogram, co-occurrence gray-level matrix (GLCM), gray-level run length matrix, gray-level zone length matrix, and neighborhood gray-level difference matrix were applied on MR images. The robustness of these features was tested in previous studies.,, [Table 3] shows the details of these features. For radiomic analysis, femoral heads were segmented in each slice on T1W, T2W, and ADC images and obtained three-dimensional (3D) regions of interest were used for feature analysis. To reduce interobserver variability, all femoral heads were segmented by an experienced radiation oncologist physicist. All segmentations were done manually and verified by an experienced radiologist (more than 10 years' experiences in musculoskeletal imaging). Before feature extraction, all image intensities were discretized to 64 gray levels, in order to image noise reduction, increasing sensitivity, and normalizing the intensities across all the patients.
All statistical analyses were performed using STATA (Version 11.1, SE Texas). Wilcoxon signed-rank test was performed to assess the significance of the change in mean T1W, T2W, and ADC radiomic features postradiotherapy relative to preradiotherapy values. Statistical significance was assumed if P < 0.05 and all reported P values are two sided. The percentage changes of the postradiotherapy to the preradiotherapy for all radiomic features were also calculated to show the relative changes postradiation. The percentage change values were normalized based on the natural logarithm base 10. This ensured that the different parameter values were in a comparable range of values when quantifying differences between pre- and postradiotherapy.
| > Results|| |
All radiomic features in T1W/T2W/ADC images had difference between pre- and postradiotherapy. Normalized percentage changes with 95% confidence intervals of these features are shown in [Figure 2], [Figure 3], [Figure 4]. We also showed the MRI radiomic features ranking in [Table 4]. Features ranked in the descending order based on the median values. High rank features showing high pre/postradiotherapy changes. Wilcoxon signed-rank test revealed that 20 T1W radiomic features vary significantly postradiotherapy. These features are shown in [Table 5]. On the contrary, ADC features did not vary significantly postradiotherapy. In regard to T2W radiomic features, only five features had significant changes postradiotherapy. The features ENTROPY_HISTOGRAM (P = 0.004), CONTRAST_GLCM (P = 0.018), ENTROPY_GLCM (P = 0.009), DISSIMILARITY_GLCM (P = 0.02), and CONTRAST_NGLDM (P = 0.013) have undergone significant changes due to radiotherapy. There was no significant correlation between mean radiation dose and radiomic feature changes. Furthermore, no fractures were observed within the follow-up time; hence, the correlation between radiomic feature changes and fracture could not be assessed.
|Figure 2: T1-weighted radiomic features – percentage Change (normalized) with 95% confidence intervals. Features with significant changes were shown with*|
Click here to view
|Figure 3: T2-weighted radiomic features – percentage change (normalized) with 95% confidence intervals. Features with significant changes were shown with*|
Click here to view
|Figure 4: Apparent diffusion coefficient radiomic features – percentage change (normalized) with 95% confidence intervals. Features with significant changes were shown with*|
Click here to view
|Table 4: Magnetic resonance imaging radiomic features ranking. Features ranked in the descending order based on median values (absolute). High rank features showing high pre/postradiation therapy changes|
Click here to view
|Table 5: T1W features which had significant changes postradiation therapy|
Click here to view
| > Discussion|| |
Radiation-induced bone complications have been observed in different patient populations undergoing radiotherapy. In a cohort study, Elliott et al. found that the risk of hip fractures significantly increases up to 76% in 45,662 Pca patients undergoing radiotherapy. In another previous retrospective and randomized prospective trials, pelvic insufficiency and hip fractures have been reported in patients treated for pelvic malignancies. In this clinical study, we demonstrated the feasibility of radiomic features to assess the radiotherapy-induced femoral head changes. We observed that all selected radiomic features have undergone changes and significant differences in 20 T1 and 5 T2 radiomic features between the pre- and postradiotherapy femoral heads.
In radiomic studies, any feature has its own concept and changes in their values could show structural and functional changes of the tissues being studied. For example, energy is associated with overall homogeneity. In our study, this feature has been changed in the postradiotherapy femoral heads, suggesting changed level of heterogeneity. Furthermore, contrast shows the difference between the highest and the lowest pixel values and is a measure of image local variations. Changes in this feature indicate that image has undergone large variations due to exposure to radiation.
A further finding of our work was that radiomic features obtained from different MR image sequences have different behaviors against radiotherapy. As shown in [Figure 2], [Figure 3], [Figure 4], some features had positive percentage changes in T1 and they had negative changes in T2 or ADC and vice versa. A possible explanation for this might be that MRI sequences (T1W/T2W/ADC) have different image acquisition mechanisms and these lead to different image features and feature response.
In recent years, researchers have investigated a variety of MRI for bone diseases detection, characterization, and classification using ADC, but to best of our knowledge, there is no report on the bone ADC feature analysis. In the present work, we showed that radiotherapy induces changes in ADC features, although statistical analysis showed that these changes are nonsignificant. It is also should be considered that b-value has a direct impact on ADC features. In our work, we used fixed b-values and in future works may use different values to assess radiotherapy-induced damages.
In regard with the role of image feature analysis for bone insufficiency, there are some interesting studies. Nardone et al. examined the role of texture analysis (TA) as a predictive factor of radiation-induced insufficiency fractures (IFs). In this study, they performed femoral head computed tomography (CT)-simulation texture analysis of patients undergoing pelvic radiotherapy and concluded that bone CT-TA could be correlated to the risk of radiation-induced IFs. They also repeated their work with (3D) TA and found that 3D-bone CT-TA can be used to stratify the risk of the patients to develop radiation-induced IFs. In another study, Rachidi et al. explored Laws' masks analysis to describe structural variations of trabecular bone due to osteoporosis on digital radiographs and found that this method constitutes a promising routine technique for the determination of osteoporosis fracture risk from radiographs. On the other hand, in a study by Thevenot et al., it was found that femoral neck fracture can be predicted on the basis of clinical radiographs using the combined analysis of bone geometry, textural analysis of trabecular bone, and bone mineral density.
Dose–volume effect for radiotherapy-induced femoral head damages has been reported in several studies and some of which indicated that there are no well-established dose constraints for femoral head. Emami et al. and Bedford et al. introduced a 52 Gy tolerance dose for the whole organ and 10% of the organ for probability of complication below 5%, respectively., In our study, despite the use of IMRT, the mean radiation dose received by femoral head was 40 Gy and there was no correlation between radiation dose and feature changes. In a recent study, Okoukoni et al. observed significant postradiotherapy cortical thinning in the intertrochanteric crest of femoral neck. They also identified that there is a correlation between radiation dose and cortical thinning.
Based on previous studies and as was reviewed by Zhang et al., due to the different responses of bone cells to radiotherapy including differentiation and radiosensitivity, the underlying mechanisms are different at different doses. However, studies have indicated that, following the radiotherapy, the initial changes in bone result from injury to the remodeling system (osteocytes, osteoblasts, and osteoclasts) which is result of direct irradiation injury to the cells of the remodeling system or the indirect result of irradiation-induced vascular injury, or a combination of both phenomena. For radiation-associated fractures, the believed mechanism is radiation-induced vascular fibrosis and impairment of osteoblast cells, which can lead to avascular necrosis and bone atrophy, therefore leaving bone susceptible to fracture. There are also some reports associating bone damages with radiation-induced cortical thinning, reduced bone mechanical properties, and reduced bone plasticity.
This is the first study to demonstrate the association between radiotherapy-induced femoral head damages and MRI radiomic feature changes and there is a lack of prior research studies on the topic. In similar works, different image texture analyses have been used to assess radiotherapy-induced shrinkage in parotid., In these studies, a little number of textural features were extracted from CT scanning and sonography and their results are specific. In comparison to these studies, we extracted several multiparametric MRI radiomic features from different feature categories and our results are different.
Although our results are original and significant, there are limitations associated with this study. First is the number of patients. Future studies will validate our findings on a larger cohort of data. Second is the time of postradiotherapy follow-up. Clinically, radiation-induced femoral head damages often take years to develop after radiotherapy and further long-term follow-up is needed to have more clinically results. However, the present study is a rapid postradiotherapy test and our results indicated that early structural change analysis using radiomic features may contribute to predict postradiotherapy fracture in these patients. Furthermore, these features can be identified as potentially important imaging biomarkers for radiotherapy-induced femoral fractures or any other complications prediction.
| > Conclusion|| |
MR radiomic feature analysis is a feasible approach for radiotherapy-induced bone damage assessment. These features may be considered as fast, noninvasive, and predictive imaging biomarkers for routine clinical use.
Financial support and sponsorship
This work was financially supported by Iran University of Medical Sciences, Grant NO. 28744.
Conflicts of interest
There are no conflicts of interest.
| > References|| |
Fiorino C, Valdagni R, Rancati T, Sanguineti G. Dose-volume effects for normal tissues in external radiotherapy: Pelvis. Radiother Oncol 2009;93:153-67.
Okoukoni C, Randolph DM, McTyre ER, Kwok A, Weaver AA, Blackstock AW, et al.
Early dose-dependent cortical thinning of the femoral neck in anal cancer patients treated with pelvic radiation therapy. Bone 2017;94:84-9.
Kwon JW, Huh SJ, Yoon YC, Choi SH, Jung JY, Oh D, et al.
Pelvic bone complications after radiation therapy of uterine cervical cancer: Evaluation with MRI. AJR Am J Roentgenol 2008;191:987-94.
Iğdem S, Alço G, Ercan T, Barlan M, Ganiyusufoğlu K, Unalan B, et al.
Insufficiency fractures after pelvic radiotherapy in patients with prostate cancer. Int J Radiat Oncol Biol Phys 2010;77:818-23.
Oh D, Huh SJ, Nam H, Park W, Han Y, Lim DH, et al.
Pelvic insufficiency fracture after pelvic radiotherapy for cervical cancer: Analysis of risk factors. Int J Radiat Oncol Biol Phys 2008;70:1183-8.
Shirazi-Fard Y, Alwood JS, Schreurs AS, Castillo AB, Globus RK. Mechanical loading causes site-specific anabolic effects on bone following exposure to ionizing radiation. Bone 2015;81:260-9.
Chandra A, Lin T, Tribble MB, Zhu J, Altman AR, Tseng WJ, et al.
PTH1-34 alleviates radiotherapy-induced local bone loss by improving osteoblast and osteocyte survival. Bone 2014;67:33-40.
Szymczyk KH, Shapiro IM, Adams CS. Ionizing radiation sensitizes bone cells to apoptosis. Bone 2004;34:148-56.
Bazzocchi A, Garzillo G, Fuzzi F, Diano D, Albisinni U, Salizzoni E, et al.
Localizer sequences of magnetic resonance imaging accurately identify osteoporotic vertebral fractures. Bone 2014;61:158-63.
Manenti G, Capuani S, Fanucci E, Assako EP, Masala S, Sorge R, et al.
Diffusion tensor imaging and magnetic resonance spectroscopy assessment of cancellous bone quality in femoral neck of healthy, osteopenic and osteoporotic subjects at 3T: Preliminary experience. Bone 2013;55:7-15.
Khoo MM, Tyler PA, Saifuddin A, Padhani AR. Diffusion-weighted imaging (DWI) in musculoskeletal MRI: A critical review. Skeletal Radiol 2011;40:665-81.
Bammer R. Basic principles of diffusion-weighted imaging. Eur J Radiol 2003;45:169-84.
Van Beers BE, Leporq B, Doblas S, Garteiser P. Imaging Biomarker Measurements. Imaging Biomarkers. Cham, Switzerland: Springer; 2017. p. 87-99.
Aerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S, et al.
Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun 2014;5:4006.
Huynh E, Coroller TP, Narayan V, Agrawal V, Hou Y, Romano J, et al.
CT-based radiomic analysis of stereotactic body radiation therapy patients with lung cancer. Radiother Oncol 2016;120:258-66.
Cunliffe A, Armato SG 3rd, Castillo R, Pham N, Guerrero T, Al-Hallaq HA, et al.
Lung texture in serial thoracic computed tomography scans: Correlation of radiomics-based features with radiation therapy dose and radiation pneumonitis development. Int J Radiat Oncol Biol Phys 2015;91:1048-56.
Pota M, Scalco E, Sanguineti G, Farneti A, Cattaneo GM, Rizzo G, et al.
Early prediction of radiotherapy-induced parotid shrinkage and toxicity based on CT radiomics and fuzzy classification. Artif Intell Med 2017;81:41-53.
Scalco E, Rizzo G. Texture analysis of medical images for radiotherapy applications. Br J Radiol 2017;90:20160642.
Abdollahi H, Mostafaei S, Cheraghi S, Shiri I, Rabi Mahdavi S, Kazemnejad A, et al.
Cochlea CT radiomics predicts chemoradiotherapy induced sensorineural hearing loss in head and neck cancer patients: A machine learning and multi-variable modelling study. Phys Med 2018;45:192-7.
Shiri I, Rahmim A, Ghaffarian P, Geramifar P, Abdollahi H, Bitarafan-Rajabi A, et al.
The impact of image reconstruction settings on 18F-FDG PET radiomic features: Multi-scanner phantom and patient studies. Eur Radiol 2017;27:4498-509.
Saeedi E, Dejkam A, Beigi J, Rastegar S, Yousef Z, Mehdipour A, et al.
Radiomic feature robustness and reproducibility in quantitative bone radiography: A study on radiologic parameter changes. J Clin Densitom. 2018. pii: S1094-6950(18)30070-2.
Shiri I, Abdollahi H, Shaysteh S, Mahdavi SR. Test-retest reproducibility and robustness analysis of recurrent glioblastoma MRI radiomics texture features. Iran J Radiol 2017;14:e48035.
Donneys A, Nelson NS, Perosky JE, Polyatskaya Y, Rodriguez JJ, Figueredo C, et al.
Prevention of radiation-induced bone pathology through combined pharmacologic cytoprotection and angiogenic stimulation. Bone 2016;84:245-52.
Elliott SP, Jarosek SL, Alanee SR, Konety BR, Dusenbery KE, Virnig BA, et al.
Three-dimensional external beam radiotherapy for prostate cancer increases the risk of hip fracture. Cancer 2011;117:4557-65.
Baxter NN, Habermann EB, Tepper JE, Durham SB, Virnig BA. Risk of pelvic fractures in older women following pelvic irradiation. JAMA 2005;294:2587-93.
Nardone V, Tini P, Carbone SF, Grassi A, Biondi M, Sebaste L, et al.
Bone texture analysis using CT-simulation scans to individuate risk parameters for radiation-induced insufficiency fractures. Osteoporos Int 2017;28:1915-23.
Nardone V, Tini P, Croci S, Carbone SF, Sebaste L, Carfagno T, et al
. 3D bone texture analysis as a potential predictor of radiation-induced insufficiency fractures. Quant Imaging Med Surg 2018;8:14-24.
Rachidi M, Marchadier A, Gadois C, Lespessailles E, Chappard C, Benhamou CL, et al.
Laws' masks descriptors applied to bone texture analysis: An innovative and discriminant tool in osteoporosis. Skeletal Radiol 2008;37:541-8.
Thevenot J, Hirvasniemi J, Pulkkinen P, Määttä M, Korpelainen R, Saarakkala S, et al.
Assessment of risk of femoral neck fracture with radiographic texture parameters: A retrospective study. Radiology 2014;272:184-91.
Emami B, Lyman J, Brown A, Coia L, Goitein M, Munzenrider JE, et al.
Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21:109-22.
Bedford JL, Khoo VS, Webb S, Dearnaley DP. Optimization of coplanar six-field techniques for conformal radiotherapy of the prostate. Int J Radiat Oncol Biol Phys 2000;46:231-8.
Zhang J, Qiu X, Xi K, Hu W, Pei H, Nie J, et al.
Therapeutic ionizing radiation induced bone loss: A review of in vivo
and in vitro
findings. Connect Tissue Res 2018;7:1-4.
Cunha SSD, Sarmento VA, Ramalho LM, Freitas AC, Almeida Dd, Tavares ME, et al
. Effects of radiotherapy on bone tissues. Radiol Bras 2007;40:189-92.
Pouya F, Kerachian MA. Avascular necrosis of the femoral head: Are any genes involved? Arch Bone Jt Surg 2015;3:149-55.
Willey JS, Lloyd SA, Nelson GA, Bateman TA. Ionizing radiation and bone loss: Space exploration and clinical therapy applications. Clin Rev Bone Miner Metab 2011;9:54-62.
Scalco E, Fiorino C, Cattaneo GM, Sanguineti G, Rizzo G. Texture analysis for the assessment of structural changes in parotid glands induced by radiotherapy. Radiother Oncol 2013;109:384-7.
Yang X, Tridandapani S, Beitler JJ, Yu DS, Yoshida EJ, Curran WJ, et al.
Ultrasound GLCM texture analysis of radiation-induced parotid-gland injury in head-and-neck cancer radiotherapy: An in vivo
study of late toxicity. Med Phys 2012;39:5732-9.
[Figure 1], [Figure 2], [Figure 3], [Figure 4]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5]