Osteoporos Int DOI 10.1007/s00198-017-3968-5
ORIGINAL ARTICLE
Bone texture analysis using CT-simulation scans to individuate risk parameters for radiation-induced insufficiency fractures V. Nardone 1 & P. Tini 1 & S. F. Carbone 2 & A. Grassi 2 & M. Biondi 3 & L. Sebaste 1 & T. Carfagno 1 & E. Vanzi 3 & G. De Otto 3 & G. Battaglia 1 & G. Rubino 1 & P. Pastina 1 & G. Belmonte 3 & L. N. Mazzoni 3 & F. Banci Buonamici 3 & M. A. Mazzei 2 & L. Pirtoli 1
Received: 10 August 2016 / Accepted: 13 February 2017 # International Osteoporosis Foundation and National Osteoporosis Foundation 2017
Abstract Summary This study deals with the role of texture analysis as a predictive factor of radiation-induced insufficiency fractures in patients undergoing pelvic radiation. Introduction This study aims to assess the texture analysis (TA) of computed tomography (CT) simulation scans as a predictive factor of insufficiency fractures (IFs) in patients with pelvic malignancies undergoing radiation therapy (RT). Methods We performed an analysis of patients undergoing pelvic RT from January 2010 to December 2014, 24 of whom had developed pelvic bone IFs. We analyzed CT-simulation images using ImageJ macro software and selected two regions of interest (ROIs), which are L5 body and the femoral head. TA parameters included mean (m), standard deviation (SD), skewness (sk), kurtosis (k), entropy (e), and uniformity (u). The IFs patients were compared (1:2 ratio) with controlled patients who had not developed IFs and matched for sex, age, menopausal status, type of tumor, use of chemotherapy, and RT dose. A reliability test of intra- and inter-reader ROI TA reproducibility with the intra-class correlation coefficient (ICC) was performed. Univariate and multivariate analyses (logistic regression) were applied for TA parameters observed both in the IFs and the controlled groups. Results Inter- and intra-reader ROI TA was highly reproducible (ICC > 0.90). Significant TA parameters on paired t test
* V. Nardone
[email protected]
1
Unit of Radiation Oncology, University Hospital of Siena, Viale Bracci, 53100 Siena, Italy
2
Unit of Diagnostic Imaging, University Hospital of Siena, Siena, Italy
3
Unit of Medical Physics, University Hospital of Siena, Siena, Italy
included L5 m (p = 0.001), SD (p = 0.002), k (p = 0.006), e (p = 0.004), and u (p = 0.015) and femoral head m (p < 0.001) and SD (p = 0.001), whereas on logistic regression analysis, L5 e (p = 0.003) and u (p = 0.010) and femoral head m (p = 0.027), SD (p = 0.015), and sex (p = 0.044). Conclusions In our experience, bone CT TA could be correlated to the risk of radiation-induced IFs. Studies on a large patient series and methodological refinements are warranted. Keywords Insufficiency fractures . Radiation therapy . Side effects . Texture analysis
Introduction Insufficiency fractures (IFs) may occur after physiologic stress to the bones with a decreased mineralization and elastic resistance and are often associated with post-menopausal or corticosteroid-induced osteoporosis [1], although the related pathogenesis is not completely understood [2, 3]. Moreover, radiation therapy (RT) to the pelvis is a risk factor for IFs of the pelvic bones and has been described after RT for gynecological [4–6], anal [7], rectal [8], and prostate cancers [9]. The incidence of post-RT IFs, although generally considered low following modern RT, is currently reported with an increasing frequency, ranging from 8.2 to 45.2% in cervical [10–12], 9 to 11.2% in rectal [8], and up to 6.8% [9] in prostate cancers. This may be due to follow-up procedures and diagnostic tools becoming more and more reliable, which may also allow for the diagnosis in oligo-symptomatic patients. However, in some cases, IFs after RT may be severely painful and invalidating, grounding the suspect of bone metastases. Risk factors for RT-induced IFs are older age, post-menopausal state, female sex, and chemotherapy [6, 8, 11, 12].
Osteoporos Int
CT-based texture analysis (TA) appreciates image heterogeneities that may not be appreciated with the naked eye, and preliminary evidence has suggested its potential value in imaging characterization for diagnostic purposes [13]. This method is based on mathematical approaches to the evaluation of gray-level intensity and position of the pixels within the image, providing the so-called Btexture features^ that represent a quantitative measure of heterogeneity [14]. To date, many studies have addressed TA in several areas of oncology, showing a potential improvement in diagnosis, characterization of images, and assessment of response to treatment [15–19]. Several investigations have dealt with bone TA, mainly for densitometry, leading to some important results, such as the development of the Trabecular Bone Score (TBS), an analytical tool for novel gray-level texture measurements on dual X-ray absorptiometry (DXA) of the lumbar spine, obtaining information related to trabecular microarchitecture [20–22]. With these premises, we investigated the potential role of a TA procedure based on CT-simulation images obtained for RT treatment planning, attempting to investigate if some of the TA results could be related to the increased risk of IFs in patients with pelvic malignancies undergoing RT.
Methods Patients Insufficiency fracture patient series From January 2010 to December 2014, 24 patients previously irradiated on pelvic content at our institution developed pelvis bone IFs during the follow-up, and their clinical, pathological, and dosimetry data is the investigation material for the present study. The insufficiency fracture patient (IF-p) series included 13 patients (54%) irradiated for anal or rectal cancer, 9 (38%) for endometrial or cervical cancer, and 2 (8%) for prostate or bladder cancer. In all cases, a CT simulation for treatment planning was obtained before RT. The presence of previous fractures of the pelvic bone was a cause of exclusion, as well as any kind of recurrence of the treated tumor in the follow-up period.
patients for sex, age, menopausal status, kind of tumor, chemotherapy administration, and RT tumor dose. The exclusion criteria were the same as for the IF-p series. In order to exclude biases due to factors not taken into account in the matched comparison, we also considered, for the comparison of the abovementioned series, the period of enrollment for RT (January 2010 to June 2012 vs. July 2012 to December 2014) and the applied RT technique (three-dimensional conformal RT (3D–CRT) vs. intensity-modulated RT (IMRT)). Radiotherapy and chemotherapy treatments RT was delivered with a linear accelerator, using 6- or 15MeV photon beams, with 3D–CRT or IMRT techniques. Target volumes and organs at risk were identified on CT simulation, with MRI image fusion if required. The mean RT doses delivered to the different pelvic bones were obtained from the dose-volume histograms resulting from treatment planning, both for IF-p and C-p groups. If deemed required according to international guidelines, chemotherapy was administered concurrently with, or sequentially to, RT, employing standard association of platinum, taxanes, mitomycin, and fluoro-pyrimidine compounds. The retrospective analysis of the data was authorized by the Internal Institutional Review Board. A signed informed consent was obtained both for any treatment and for the anonymous use of clinical data. All procedures were undertaken in compliance with the ethical statements of the Helsinki Declaration (1964, amended most recently in 2008) of the World Medical Association. Follow-up After completion of RT, all patients entered a scheduled follow-up program according to tumor type. In patients affected by gynecological, gastrointestinal, and bladder malignancies, a CT and/or MRI examination was obtained at 4 and 12– 16 weeks after RT, then every 6 months or in any case showing clinical signs suggesting tumor recurrence. In prostate cancer patients, diagnostic CT and/or MRI was obtained only if justified by a rise of the PSA value and/or by emerging symptoms (including pain) or physical signs of recurrence or complications. A physical examination, blood counts, and chemistry were obtained every 3 months.
Controlled patient series Assessment of IFs The above cohort of patients was evaluated through a matched analysis of the corresponding data of 48 patients who were also treated in our institution. The IF-p series was compared (1:2 ratio) with the controlled patient (C-p) series, composed of patients submitted to pelvic irradiation in our institution but without developing IFs, matching any IF-p with two C-p
The occurrence of IFs was confirmed at CT or MR examinations, performed during the follow-up, by a radiologist (SFC) with 14-year experience in the oncologic field. CT findings of IF were defined as fracture lines or sclerotic linear changes in the bones, and MRI findings of IFs were defined as signal
Osteoporos Int
intensity changes in the bones of >5 mm both on T1- and T2weighted images [6]. The simulation CT was reviewed for pre-existent fractures, and if a fracture was identified on the simulation CT, the patient was excluded from the study. Patients developing any kind of tumor recurrence were not included in the present evaluation, as previously stated.
CT simulation CT simulation was performed upon referral for RT, with a 16slice CT-simulation scanner (General Electric, Boston, USA, LightSpeed, RT16, slice thickness 2.5 mm, beam pitch of 1.375, reconstruction interval 2.5 mm, tube voltage of 120 kVp and reference mAs ranging from 100 to 440 mA, noise index 10). The stability of the CT to electron density table was annually validated with the CIRS Phantom model 62. The ICRP 162 tolerance was applied (<20 HU from the baseline), and in the time period considered (January 2010 to December 2014), the values were stable.
Image analysis We selected two regions of interest (ROIs) on CT simulation, being the vertebral body of L5 and the femoral head (Fig. 1). Each ROI was independently contoured by a radiation oncologist (VN) twice for each case with a 24–48-h interval and by a resident in radiology (AG), once for each case, keeping the contour as close as possible to the internal side of the compact bone. For the vertebral body of L5, the cranio-caudal anatomical landmark was the middle of the body, whereas for the femoral head, the cranio-caudal landmark was the fovea capitis (Fig. 1). The TA was accomplished using a homemade ImageJ macro, implementing a first-order, statistical-based method. With this kind of analysis, considering the gray-level frequency distribution obtained from the histogram of pixel intensities, it is possible to take into account the distribution of gray-level values in each ROI (Fig. 1). This approach is considered as the first order; in that, it is dependent on single-pixel values rather than pixel-topixel interactions. The following typical TA parameters were evaluated: mean (m), standard deviation (SD), skewness (sk), kurtosis (k), entropy (e), and uniformity (u). In Fig. 2, we report the mathematical functions implemented to calculate each parameter used in this study. In particular, m and SD refer to the mean intensity and the standard deviation of the values in the histogram of pixel intensities, sk to the asymmetry, and k to the flatness of the histogram, whereas u and e are, respectively, parameters of uniformity and of irregularity of gray-level distribution.
Fig. 1 Examples of ROI and histograms of the pixel distribution (CTsimulation DICOM images). a L5 ROI in a C-p patient. b L5 ROI in an IF-p patient. c Femoral head ROI in a C-p patient. d Femoral head ROI in an IF-p patient
Statistics Inter- and intra-reader reproducibility of the results were assessed using the intra-class correlation coefficient (ICC) method [23]. The TA parameters, also, were tested for co-correlation (Pearson’s correlation). The TA parameters were correlated with the development of IFs by paired t test (each patient in IF-p against his matched patients in C-p averaged). Binary logistic regression was tested with group as dependant variable and all the texture parameters plus all relevant confounding variables (sex, radiation doses, chemotherapy, RT technique, enrollment period), with forward conditional method, after standard deviation normalization of the TA parameters. The resulted odds ratio (OR) was then normalized. The mean doses delivered to the different pelvic bones in IF-p vs. C-p groups were compared using the Mann-Whitney test. We considered as significant a p value <0.005. The statistical analysis was performed using the SPSS software 17.0.
Osteoporos Int Fig. 2 Mathematical functions used for TA parameter calculations
Results The characteristics of IF-p and C-p series, as well as the localization of the Ifs, are reported in Table 1. IFs occurred in the sacroiliac joints, pubis, acetabulum, sacral body, and lumbar Table 1 Characteristics of patients in the insufficiency fracture (IF-p) and control (C-p) series
vertebrae. Fifteen patients (58%) developed multiple IFs. The median follow-up time was 47.34 months (mean 49.24 months, SD 22.12 months, range 14–76 months). Regarding the period of enrollment, 10 out of 24 patients (41%) in IF-p series and 22 out of 48 patients (46%) in C-p
Characteristics of patients
IF-p series
C-p series
Sex
6 (25%)
12 (25%)
18 (75%)
36 (75%)
66.5 years (±10.99 years)
63.5 years (±12.17 years)
Range 30–81 years
Range 32–78 years
5 (27%)
10 (27%)
13 (73%)
26 (73%)
Male Female Age Menopausal status
Pre-menopausal Post-menopausal Disease
9 (38%)
18 (38%)
Gynecological
13 (54%)
26 (54%)
Gastrointestinal
2 (8%)
4 (8%)
Yes 16 (66%)
Yes 32 (66%)
No 8 (34%)
No 16 (34%)
Mean 5030 cGy (±480 cGy)
Mean 5060 cGy (±530 cGy)
Range 4500–5940 cGy
Range 4500–5940 cGy
Urological Chemotherapy RT target dose (PTV) Localization of the IFs
14 (58%)
Sacroiliac joints
7 (29%)
Pubis
2 (8%)
Acetabulum
5 (20%)
Sacral body
4 (16%)
Lumbar vertebrae Enrollment period
14/24 (59%)
26/48 (54%)
10/24 (41%)
22/48 (46%)
RT technique
13/24 (54%)
28/48 (58%)
IMRT
11/24 (46%)
20/48 (42%)
January 2010 to June 2012 July 2012 to December 2014
3D–CRT
Osteoporos Int
IF-p
Fig. 3 Analysis of mean RT doses delivered to pelvic bones (Mann-Whitney test) in IF-p and C-p patients
C-p
p-value
0,718
L5 10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
0,198
Sacrum 10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
0,355
Ilium
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
Ischium
0,588 10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
0,924
Pubis
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
Femoral Head
0,456
10 Gy
series were enrolled from January 2010 to June 2012, whereas 14 out of 24 patients (59%) in IF-p series and 26 out of 48 patients (54%) in C-p series were enrolled from July 2012 to December 2014. Regarding the RT technique, 13 out of 24 patients (54%) in IF-p series and 28 out of 48 patients (58%) in C-p series underwent 3D–CRT, whereas 11 out of 24 patients (46%) in IF-p series and 20 out of 48 patients (42%) in C-p series underwent IMRT. As resulting in dose-volume
20 Gy
30 Gy
40 Gy
50 Gy
10 Gy
20 Gy
30 Gy
40 Gy
50 Gy
histograms (p values in the 0.198–0.924 range), the doses on the pelvic bones between IF-p and C-p were well balanced between the groups (see the results reported in Fig. 3). Test reproducibility After the definition of the window level in Hounsfield unit (level 200, window 1000) and of the spatial definition of the
Osteoporos Int Table 2 Reproducibility (statistical analysis, ICC; see text) of TA parameters between the IF-p and C-p series Texture analysis parameter
Intra-reader ICC
Inter-reader ICC
L5 mean
0.924
0.923
L5 standard deviation
0.936
0.918
L5 skewness L5 kurtosis
0.942 0.921
0.874 0.894
L5 entropy
0.907
0.854
L5 uniformity Femoral head mean
0.901 0.945
0.864 0.974
Femoral head standard deviation Femoral head skewness
0.942 0.938
0.938 0.981
Femoral head kurtosis
0.952
0.948
Femoral head entropy
0.932
0.912
Femoral head uniformity
0.913
0.923
ROI (the inner aspect of the cortical bone at middle L5 and at the level of fovea in femoral head), we obtained a high reproducibility of the method, with an inter- and intra-reader ICC > 85% for each evaluated parameter (Table 2). Texture analysis parameters related to the development of IFs after matched comparison The TA parameters that resulted significant at the paired t test analysis were L5 m (p = 0.001), SD (p = 0.002), k (p = 0.006), e (p = 0.004), and u (p = 0.015) and femoral head m (p < 0.001) and SD (p = 0.001) (Table 3). On binary logistic regression analysis with group as dependant variable and all the texture parameters plus all relevant confounding variables (sex, radiation doses, chemotherapy, RT technique, enrollment period), the variables that resulted significant were L5 e (p = 0.003, OR 1.24) and u (p = 0.010, OR 1.87) and femoral head m (p = 0.027, OR 1.23), SD (p = 0.015, OR 1.24), and the sex (p = 0.044, OR 2.81), with a R2 0.672 (Table 3). The TA parameters, finally, were tested for co-correlation with Pearson’s correlation analysis (Table 4).
Discussion Radiation-induced bone damage (and in particular IFs) has been described since the beginning of the twentieth century, and moreover, the incidence of IFs has apparently increased in recent observations, probably due to more thorough follow-up policies including sophisticated imaging resources, although there remain many discrepancies in various studies [8–12]. One MRI study reported an
impressive (89%) incidence of findings compatible with IFs from patients who had previously undergone pelvic RT [24], while a bone nuclide scan could detect a 34% incidence [25]. Although the majority of the IFs develop within 1 year after the RT [24], the temporal limit is not defined [26]. Therefore, our follow-up, with a median value of nearly 4 years, can be adequate. Several authors have recently addressed the importance of high RT doses to bone, advanced age, BMD loss, and chemotherapy [27–30], even with the development of a post-RT, BMD-based predictive model for IFs [29]. Very recently [20–22], the TBS measurement of gray-level texture on DXA images provided information on trabecular microarchitecture of the bone, and this seems consistently associated with both prevalent and incident fractures, partially independent from both clinical risk factors and BMD. At this regard, patients undergoing RT usually do not undergo DXA, so we chose to apply a similar method of imaging analysis to the simulation CT. We chose to apply a Bsingle-slice^ analysis and to focalize on the parameters known to be correlated with the bone density and microarchitecture; therefore, we chose a limited number of texture parameters, as previously published in a short version of this work [31]. Only Uezono et al. [32] have taken into account pre-RT, CT imaging-detectable parameters (i.e., low bone density) as predictive factors for IFs, but their data are not robust. Our method of analysis, using a matched comparison, allows to focalize on the texture parameters and does not take into account the radiation therapy doses, or the other parameters known to be related to IFs (i.e., age, chemotherapy, sex), and this is demonstrated by their non-significance in logistic regression analysis, except for the sex, whose significance is due to the ratio between male and females in the patients series. In fact, a significant characterization was shown in terms of TA parameters of L5 and femoral heads upon the univariate and multivariate analyses. The femoral head TA parameters m and SD, as well as the L5 TA parameters SD and e, were significantly lower in the IFp series than in C-p series, whereas the L5 TA parameters k and u were significantly higher in the IF-p series than in C-p series. We can speculate, at this regard, that the lower m reflects a lower density and a lower mineralization of the bone, in accordance to other works in the literature [20, 33]. The lower SD and the higher k, as well as the lower e and the higher u, at the same time, are correlated to the trabecular microarchitecture, in accordance with the work of Rachidi et al. [34] and Thevenot et al. [35]. In their studies, patients with IFs had lower entropy than the C-p, and this might be
Osteoporos Int Table 3 Paired t test analysis and binary logistic regression analysis
Paired t test analysis Paired differences (IF-p) − (C-p) TA parameters Mean
L5
Mean −42.53 ± 44.62
95% CI −61.37/−23.68
p value 0.001
−71.13 ± 64.55
−98.39/−43.88
<0.001
Femoral head Standard deviation
L5
−23.86 ± 32.96
−37.78/−9.94
0.002
−16.76 ± 20.93
−25.61/−7.92
0.001
Femoral head Skewness
0.26 ± 0.78
−0.068/0.595
0.115
L5
−0.05 ± 0.73
−0.367/0.254
0.710
Femoral head Kurtosis
2.21 ± 3.60
0.69/3.73
0.006
L5
−0.61 ± 3.48
−2.08/0.85
0.393
Entropy
−0.21 ± 0.34
−0.363/−0.075
0.004
L5
0.101 ± 0.309
−0.029/0.232
0.121
0.0018 ± 0.0034
0.00038/0.00328
0.015
−0.00062 ± 0.00229
−0.0015/0.00034
0.194
Femoral head
Femoral head Uniformity
L5 Femoral head
Logistic regression analysis (normalized odds ratio) Parameter p value Entropy (L5) 0.003 Uniformity (L5) 0.010 Mean (FH) 0.027 Standard deviation (FH) 0.015 Sex 0.044
Odds ratio 1.24 (protection) 1.87 (risk) 1.23 (protection) 1.24 (protection) 2.81 (risk for F)
95% CI 1.08–1.35 1.34–2.64 1.02–1.49 1.04–1.48 1.06–7.48
The normalized odds ratios were divided in Bprotection^ and Brisk^ according to their positive or negative correlation with the outcome
explained by the fibers being more marked in the control group, with an increase in the randomness of the pixel values, and eventually an increase in the entropy and a decrease in uniformity. Table 4
However, some aspects are worthy of consideration. This analysis intended to provide a reliable and reproducible method, suitable for identifying patients who are prone to develop pelvic RT-induced IFs prior to treatment.
Co-correlation between TA parameters (Pearson’s correlation analysis)
L5 mean L5 standard deviation L5 skewness L5 kurtosis L5 entropy L5 uniformity FH mean FH standard deviation FH skewness FH kurtosis FH entropy FH uniformity
L5 L5 standard mean deviation
L5 L5 L5 L5 FH skewness kurtosis entropy uniformity mean
FH standard deviation
FH FH FH FH skewness kurtosis entropy uniformity
1 0.776 0.776 1
0.238 0.059
0.367 0.167
0.248 0.172
0.221 0.087
0.544 0.021 0.422 0.002
0.039 0.056
0.038 0.080
0.398 0.339
0.348 0.306
0.248 0.059 0.367 0.167
1 0.917
0.917 1
0.837 0.900
0.893 0.906
0.194 0.032 0.273 0.067
0.181 0.135
0.131 0.060
0.013 0.074
0.018 0.061
0.248 0.221 0.544 0.021
0.172 0.087 0.422 0.002
0.837 0.893 0.194 0.032
0.900 0.906 0.273 0.067
1 0.974 0.120 0.020
0.974 1 0.106 0.008
0.120 0.106 1 0.464
0.020 0.008 0.464 1
0.063 0.089 0.332 0.078
0.012 0.017 0.177 0.214
0.066 0.022 0.084 0.036
0.067 0.033 0.045 0.001
0.039 0.038 0.398 0.348
0.056 0.080 0.339 0.306
0.181 0.131 0.013 0.018
0.135 0.060 0.074 0.061
0.063 0.012 0.066 0.067
0.089 0.017 0.022 0.033
0.332 0.177 0.084 0.045
0.078 0.214 0.036 0.001
1 0.898 0.723 0.779
0.898 1 0.804 0.871
0.723 0.804 1 0.986
0.779 0.871 0.986 1
Osteoporos Int
The ROI contouring was done by a radiation oncologist and by a radiologist, and the learning curve was minimum, due to their knowledge of this kind of analysis. The definition of the ROI (window level, anatomical landmark, and contouring as close as possible to the internal side of the compact bone) was critical to achieve the reliability results, but at the same time, we did not get information regarding the cortical part of the bones, whose contouring was not reproducible. We chose the CT simulation due to the uniformity of the scanner parameters in the patient series, as well as for the phantom-based calibration, regularly performed. Finally, the interpretation of TA parameters on the grounds of bone pathophysiology is incomplete, also, resulting at present by related literature [33–35]; thus, our study and other similar ones are still awaiting an exhaustive scientific background.
anonymous use of clinical data. All procedures were undertaken in compliance with the ethical statements of the Helsinki Declaration (1964, amended most recently in 2008) of the World Medical Association. Conflicts of interest None.
References 1.
2.
3.
4.
Limitations of the study 5.
Our results may be worthy of critical consideration for possible methodological and technical refinements and a desirable more fitting correlation with post-RT IFs. In particular, the single-slice analysis, in respect of the volumetric analysis, is more prone to errors due to the operator and, at the same time, can produce only a limited number of texture parameters [36–38]. At this regard, our choice of the TA parameters could be complemented by other TA parameters, such as the TBS and its precursor variogram slope that has been successfully used for TA [20, 22]. The choice of the matched comparison, also, can overestimate the effects of the TA parameters in respect of the other known variables (i.e., radiation doses, sex, age, menopausal state), so the effective value of the TA parameters needs to be investigated in a prospective study to substantiate this field of investigation. Further, we need to investigate the real reproducibility and the reliability of this kind of analysis in other departments and hospitals, with different parameters of CT acquisitions and different specialists.
Conclusions Despite the limitations of the study, the reported data are nevertheless worthy of attention. At our knowledge, these data are original, as no one correlated bone CT TA to the risk of radiation-induced IFs. A prospective analysis is needed to validate and estimate the reliability of this kind of approach. Compliance with ethical standards The retrospective analysis of the data was authorized by the Internal Institutional Review Board. A signed informed consent was obtained both for any treatment and for the
6.
7.
8.
9.
10.
11.
12.
Lundin B, Bjorkholm E, Lundell M, Jacobsson H (1990) Insufficiency fractures of the sacrum after radiotherapy for gynaecological malignancy. Acta Oncol 29(2):211–215 Pacheco R, Stock H (2013) Effects of radiation on bone. Current osteoporosis reports 11(4):299–304. doi:10.1007/ s11914-013-0174-z Fajardo LF (2005) The pathology of ionizing radiation as defined by morphologic patterns. Acta Oncol 44(1):13–22. doi:10.1080/ 02841860510007440 Huh SJ, Kim B, Kang MK, Lee JE, Lim DH, Park W, Shin SS, Ahn YC (2002) Pelvic insufficiency fracture after pelvic irradiation in uterine cervix cancer. Gynecol Oncol 86(3):264–268 Shih KK, Folkert MR, Kollmeier MA, Abu-Rustum NR, Sonoda Y, Leitao MM Jr, Barakat RR, Alektiar KM (2013) Pelvic insufficiency fractures in patients with cervical and endometrial cancer treated with postoperative pelvic radiation. Gynecol Oncol 128(3):540– 543. doi:10.1016/j.ygyno.2012.12.021 Tokumaru S, Toita T, Oguchi M, Ohno T, Kato S, Niibe Y, Kazumoto T, Kodaira T, Kataoka M, Shikama N, Kenjo M, Yamauchi C, Suzuki O, Sakurai H, Teshima T, Kagami Y, Nakano T, Hiraoka M, Mitsuhashi N, Kudo S (2012) Insufficiency fractures after pelvic radiation therapy for uterine cervical cancer: an analysis of subjects in a prospective multiinstitutional trial, and cooperative study of the Japan Radiation Oncology Group (JAROG) and Japanese Radiation Oncology Study Group (JROSG). Int J Radiat Oncol Biol Phys 84(2):e195– e200. doi:10.1016/j.ijrobp.2012.03.042 Allal AS, Mermillod B, Roth AD, Marti MC, Kurtz JM (1997) Impact of clinical and therapeutic factors on major late complications after radiotherapy with or without concomitant chemotherapy for anal carcinoma. Int J Radiat Oncol Biol Phys 39(5):1099–1105 Kim HJ, Boland PJ, Meredith DS, Lis E, Zhang Z, Shi W, Yamada YJ, Goodman KA (2012) Fractures of the sacrum after chemoradiation for rectal carcinoma: incidence, risk factors, and radiographic evaluation. Int J Radiat Oncol Biol Phys 84(3):694–699. doi:10. 1016/j.ijrobp.2012.01.021 Igdem S, Alco G, Ercan T, Barlan M, Ganiyusufoglu K, Unalan B, Turkan S, Okkan S (2010) Insufficiency fractures after pelvic radiotherapy in patients with prostate cancer. Int J Radiat Oncol Biol Phys 77(3):818–823. doi:10.1016/j.ijrobp.2009.05.059 Ogino I, Okamoto N, Ono Y, Kitamura T, Nakayama H (2003) Pelvic insufficiency fractures in postmenopausal woman with advanced cervical cancer treated by radiotherapy. Radiother Oncol 68(1):61–67 Ikushima H, Osaki K, Furutani S, Yamashita K, Kishida Y, Kudoh T, Nishitani H (2006) Pelvic bone complications following radiation therapy of gynecologic malignancies: clinical evaluation of radiation-induced pelvic insufficiency fractures. Gynecol Oncol 103(3):1100–1104. doi:10.1016/j.ygyno.2006.06.038 Oh D, Huh SJ, Nam H, Park W, Han Y, Lim do H, Ahn YC, Lee JW, Kim BG, Bae DS, Lee JH (2008) Pelvic insufficiency fracture after pelvic radiotherapy for cervical cancer: analysis of risk factors. Int J Radiat Oncol Biol Phys 70(4):1183–1188. doi:10.1016/j. ijrobp.2007.08.005
Osteoporos Int 13.
Davnall FYCSP, Ljungqvist G et al (2012) Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insights Imaging 3:573–589 14. Kato H, Kanematsu M, Zhang X, Saio M, Kondo H, Goshima S, Fujita H (2007) Computer-aided diagnosis of hepatic fibrosis: preliminary evaluation of MRI texture analysis using the finite difference method and an artificial neural network. AJR Am J Roentgenol 189(1):117–122. doi:10.2214/AJR.07.2070 15. Ganeshan B, Skogen K, Pressney I, Coutroubis D, Miles K (2012) Tumour heterogeneity in oesophageal cancer assessed by CT texture analysis: preliminary evidence of an association with tumour metabolism, stage, and survival. Clin Radiol 67(2):157–164. doi: 10.1016/j.crad.2011.08.012 16. Ganeshan BAS, Young RC, Chatwin CR, Miles KA (2010) Texture analysis of non-small cell lung cancer on unenhanced computed tomography: initial evidence for a relationship with tumour glucose metabolism and stage. Cancer Imaging 10:137–143 17. Huang YL, Chen JH, Shen WC (2006) Diagnosis of hepatic tumors with texture analysis in nonenhanced computed tomography images. Acad Radiol 13(6):713–720. doi:10.1016/j.acra.2005.07.014 18. Cui C, Cai H, Liu L, Li L, Tian H, Li L (2011) Quantitative analysis and prediction of regional lymph node status in rectal cancer based on computed tomography imaging. Eur Radiol 21(11):2318–2325. doi:10.1007/s00330-011-2182-7 19. Goh V, Ganeshan B, Nathan P, Juttla JK, Vinayan A, Miles KA (2011) Assessment of response to tyrosine kinase inhibitors in metastatic renal cell cancer: CT texture as a predictive biomarker. Radiology 261(1):165–171. doi:10.1148/radiol.11110264 20. Harvey NC, Gluer CC, Binkley N, McCloskey EV, Brandi ML, Cooper C, Kendler D, Lamy O, Laslop A, Camargos BM, Reginster JY, Rizzoli R, Kanis JA (2015) Trabecular Bone Score (TBS) as a new complementary approach for osteoporosis evaluation in clinical practice. Bone 78:216–224. doi:10.1016/j.bone. 2015.05.016 21. Bousson V, Bergot C, Sutter B, Thomas T, Bendavid S, Benhamou CL, Blain H, Brazier M, Breuil V, Briot K, Chapurlat R, Chapuis L, Cohen Solal M, Fardellone P, Feron JM, Gauvain JB, Laroche M, Legrand E, Lespessailles E, Linglart A, Marcelli C, Roux C, Souberbielle JC, Tremollieres F, Weryha G, Cortet B, Groupe de Recherche et d’Information sur les O (2015) Trabecular Bone Score: where are we now? Joint, bone, spine : revue du rhumatisme 82(5):320–325. doi:10.1016/j.jbspin.2015.02.005 22. Di Gregorio S, Del Rio L, Rodriguez-Tolra J, Bonel E, Garcia M, Winzenrieth R (2015) Comparison between different bone treatments on areal bone mineral density (aBMD) and bone microarchitectural texture as assessed by the Trabecular Bone Score (TBS). Bone 75:138–143. doi:10.1016/j.bone.2014.12.062 23. Ionan AC, Polley MY, McShane LM, Dobbin KK (2014) Comparison of confidence interval methods for an intra-class correlation coefficient (ICC). BMC Med Res Methodol 14:121. doi: 10.1186/1471-2288-14-121 24. Blomlie V, Rofstad EK, Talle K, Sundfor K, Winderen M, Lien HH (1996) Incidence of radiation-induced insufficiency fractures of the female pelvis: evaluation with MR imaging. AJR Am J Roentgenol 167(5):1205–1210. doi:10.2214/ajr.167.5.8911181 25. Abe H, Nakamura M, Takahashi S, Maruoka S, Ogawa Y, Sakamoto K (1992) Radiation-induced insufficiency fractures of the pelvis: evaluation with 99mTc-methylene diphosphonate
scintigraphy. AJR Am J Roentgenol 158(3):599–602. doi:10. 2214/ajr.158.3.1739002 26. Baxter NN, Habermann EB, Tepper JE, Durham SB, Virnig BA (2005) Risk of pelvic fractures in older women following pelvic irradiation. JAMA 294(20):2587–2593. doi:10.1001/jama.294.20. 2587 27. Otani K, Teshima T, Ito Y, Kawaguchi Y, Konishi K, Takahashi H, Ohigashi H, Oshima K, Araki N, Nishiyama K, Ishikawa O (2016) Risk factors for vertebral compression fractures in preoperative chemoradiotherapy with gemcitabine for pancreatic cancer. Radiother Oncol 118(3):424–429. doi:10.1016/j.radonc.2016.01. 006 28. Uyterlinde W, Chen C, Belderbos J, Sonke JJ, Lange C, de Bois J, van den Heuvel M (2016) Fractures of thoracic vertebrae in patients with locally advanced non-small cell lung carcinoma treated with intensity modulated radiotherapy. Radiother Oncol 118(3):437– 441. doi:10.1016/j.radonc.2015.11.011 29. Wei RL, Jung BC, Manzano W, Sehgal V, Klempner SJ, Lee SP, Ramsinghani NS, Lall C (2016) Bone mineral density loss in thoracic and lumbar vertebrae following radiation for abdominal cancers. Radiother Oncol 118(3):430–436. doi:10.1016/j.radonc.2016. 03.002 30. Pilz K, Hoffmann AL, Baumann M, Troost EG (2016) Vertebral fractures—an underestimated side-effect in patients treated with radio(chemo)therapy. Radiother Oncol 118(3):421–423. doi:10. 1016/j.radonc.2016.02.021 31. Nardone V, Tini P, Sebaste L, Biondi M, Banci Buonamici F, Pirtoli L (2016) Bone structure texture analysis as a potential tool to estimate radiation induced insufficiency fracture risk. Radiother Oncol. doi:10.1016/j.radonc.2016.06.001 32. Uezono H, Tsujino K, Moriki K, Nagano F, Ota Y, Sasaki R, Soejima T (2013) Pelvic insufficiency fracture after definitive radiotherapy for uterine cervical cancer: retrospective analysis of risk factors. J Radiat Res 54(6):1102–1109. doi:10.1093/jrr/rrt055 33. Christopher JJ, Ramakrishnan S (2008) Assessment and classification of mechanical strength components of human femur trabecular bone using texture analysis and neural network. J Med Syst 32(2): 117–122 34. Rachidi M, Marchadier A, Gadois C, Lespessailles E, Chappard C, Benhamou CL (2008) Laws’ masks descriptors applied to bone texture analysis: an innovative and discriminant tool in osteoporosis. Skelet Radiol 37(6):541–548. doi:10.1007/s00256-008-0463-2 35. Thevenot J, Hirvasniemi J, Pulkkinen P, Maatta M, Korpelainen R, Saarakkala S, Jamsa T (2014) Assessment of risk of femoral neck fracture with radiographic texture parameters: a retrospective study. Radiology 272(1):184–191. doi:10.1148/radiol.14131390 36. Galavis PEHC, Jallow N, Paliwal B, Jeraj R (2010) Variability of textural features in FDG PET images due to different acquisition modes and reconstruction parameters. Acta Oncol 49(7):1012– 1016 37. Waugh SALR, Bidaut L, Thompson AM (2011) The influence of field strength and different clinical breast MRI protocols on the outcome of texture analysis using foam phantoms. Med Phys 38(9):5058–5066 38. Collewet GSM, Mariette F (2004) Influence of MRI acquisition protocols and image intensity normalization methods on texture classification. Magn Reson Imaging 22(1):81–91