CME JOURNAL OF MAGNETIC RESONANCE IMAGING 39:1049–1078 (2014)

CME Article

Therapy Monitoring of Skeletal Metastases With Whole-Body Diffusion MRI Anwar R. Padhani, MB, BS1*, Andreas Makris, MD2, Peter Gall, PhD3, David J. Collins, BA4, Nina Tunariu, MB, BS4, and Johann S. de Bono, MB, BS, PhD5 This article is accredited as a journal-based CME activity. If you wish to receive credit for this activity, please refer to the website: www.wileyhealthlearning.com/jmri ACCREDITATION AND DESIGNATION STATEMENT Blackwell Futura Media Services designates this journal based CME activity for a maximum of 1 AMA PRA Category 1 CreditTM. Physicians should only claim credit commensurate with the extent of their participation in the activity. Blackwell Futura Media Services is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. EDUCATIONAL OBJECTIVES Upon completion of this educational activity, participants will be better able to: 1. Recognize the response of metastatic bone disease using diffusion weighted MRI. 2.Identify the causes of false positives and false negative results when applying diffusion-weighted MRI for bone lesion detection. ACTIVITY DISCLOSURES No commercial support has been accepted related to the development or publication of this activity. Faculty Disclosures: Editor-in-Chief: Mark E. Schweitzer, MD, discloses DSMB work for Paradigm Spine, and consultation for MMI. CME Editor: Scott B. Reeder, MD, PhD has no conflicts of interest to disclose. CME Committee: Pratik Mukherjee, MD, PhD, Shreyas Vasanawala, MD, PhD, Bonnie Joe, MD, PhD, Tim Leiner, MD, PhD, Sabine Weckbach, MD, and Frank Korosec, PhD have no conflicts of interest to disclose. Scott K. Nagle, MD, PhD discloses a personal shareholder investment in GE. Mustafa R. Bashir, MD discloses research support from Bracco Diagnostics and Siemens HealthCare, and consultant honorarium from Bayer Pharmaceuticals.

Authors: Anwar R. Padhani, MB, BS, discloses speaker honoraria from Siemens HealthCare. Andreas Makris, MD; Peter Gall, PhD; David J. Collins, BA; Nina Tunariu, MB, BS; and Johann S. de Bono, MB, BS, PhD, have no relevant financial relationships to disclose. This manuscript underwent peer review in line with the standards of editorial integrity and publication ethics maintained by Journal of Magnetic Resonance Imaging. The peer reviewers have no relevant financial relationships. The peer review process for Journal of Magnetic Resonance Imaging is double-blinded. As such, the identities of the reviewers are not disclosed in line with the standard accepted practices of medical journal peer review. Conflicts of interest have been identified and resolved in accordance with Blackwell Futura Media Services’ Policy on Activity Disclosure and Conflict of Interest. INSTRUCTIONS ON RECEIVING CREDIT For information on applicability and acceptance of CME credit for this activity, please consult your professional licensing board. This activity is designed to be completed within an hour; physicians should claim only those credits that reflect the time actually spent in the activity. To successfully earn credit, participants must complete the activity during the valid credit period. Follow these steps to earn credit:  Log on to www.wileyhealthlearning.com  Read the target audience, educational objectives, and activity disclosures.  Read the article in print or online format.  Reflect on the article.  Access the CME Exam, and choose the best answer to each question.  Complete the required evaluation component of the activity. This activity will be available for CME credit for twelve months following its publication date. At that time, it will be reviewed and potentially updated and extended for an additional period.

1

Paul Strickland Scanner Centre, Mount Vernon Hospital, Northwood, Middlesex, UK Mount Vernon Cancer Centre, Mount Vernon Hospital, Northwood, Middlesex, UK 3 Siemens AG, Healthcare Sector, Imaging & Therapy Division, Magnetic Resonance, H IM MR PLM-AW ONCO, Erlangen, Germany 4 Cancer Research UK Clinical Magnetic Resonance Research Group, Institute of Cancer Research and Royal Marsden NHS Foundation Trust, Sutton, Surrey, UK 5 Drug Development Unit, Institute of Cancer Research and Royal Marsden NHS Trust, Sutton, Surrey, UK. *Address reprint requests to: A.R.P., Paul Strickland Scanner Centre, Mount Vernon Hospital, Rickmansworth Road, Northwood, Middlesex, HA6 2RN, UK. E-mail: [email protected] Received August 28, 2012; Accepted July 25, 2013. DOI 10.1002/jmri.24548 View this article online at wileyonlinelibrary.com. 2

C 2014 Wiley Periodicals, Inc. V

1049

1050 Current methods of assessing tumor response at skeletal sites with metastatic disease use a combination of imaging tests, serum and urine biochemical markers, and symptoms assessment. These methods do not always enable the positive assessment of therapeutic benefit to be made but instead provide an evaluation of progression, which then guides therapy decisions in the clinic. Functional imaging techniques such as whole-body diffusion magnetic resonance imaging (MRI) when combined with anatomic imaging and other emerging “wet” biomarkers can improve the classification of therapy response in patients with metastatic bone disease. A range of imaging findings can be seen in the clinic depending on the type of therapy and duration of treatment. Successful response to systemic therapy is usually depicted by reductions in signal intensity accompanied by apparent diffusion coefficient (ADC) increases. Rarer patterns of successful treatment include no changes in signal intensity accompanying increases in ADC values (T2 shine-through pattern) or reductions in signal intensity without ADC value changes. Progressive disease results in increases in extent/intensity of disease on high b-value images with variable ADC changes. Diffusion MRI therapy response criteria need to be developed and tested in prospective studies in order to address current, unmet clinical and pharmaceutical needs for reliable measures of tumor response in metastatic bone disease. Keywords: bone metastases; diffusion MRI; therapy monitoring; whole-body MRI; breast and prostate cancer J. Magn. Reson. Imaging 2014;39:1049–1078. C 2014 Wiley Periodicals, Inc. V

METASTATIC BONE DISEASE is a common manifestation of advanced cancers, with autopsy studies indicating a very high prevalence in breast, prostate (more than 70%), and lung cancers (1–3). In many patients who are originally diagnosed with breast or prostate cancers, the bulk of the tumor burden at the time of death will be in the skeleton. Osteolytic metastases in particular cause bone and compressive nerve pain, impairs mobility, results in pathological fractures, and causes spinal cord compression. Other consequences of bone metastases include anemia and symptoms related to hypercalcemia. Thus, the presence of metastases in bone is a poor prognostic sign, with low survival rates of a few months for patients with lung cancer but survival can be prolonged for hormone receptor-positive breast and prostate cancers (3). On imaging, there is a spectrum of bone findings ranging from mostly destructive or osteolytic (exemplified by breast cancer) to mostly bone-forming or osteoblastic (often seen with prostate cancer), which have differing biologic mechanisms (4,5). Osteolysis is caused by osteoclast stimulation and not by the direct effects of cancer cells on bone (6). Although the dominant lesion is destructive, there is usually local bone formation which likely represents attempts at bone repair (7). Osteoblastic metastases result from profound local stimulation of osteoblasts by metastatic tumor cells present within the adjacent bone marrow. It is important to recognize that the processes of bone resorption and bone formation are almost always linked in health, but in cancer this coupling is both elevated and markedly distorted. Thus, in the spectrum of bone metastasis, predominantly osteolytic lesions

Padhani et al.

are associated with relatively reduced osteoblastic activity and predominantly osteoblastic lesions have relatively reduced resorption components. Regardless of the pattern of bone metastases, it is clear that there is elevated bone turnover (both bone resorption and formation) compared to the normal state. Once bony metastases have occurred, cancer cure becomes impossible and therapy is instituted with a palliative intent. Therapy goals are to alleviate symptoms, improve quality of life, delay progression, and achieve a modest survival benefit if possible. Bone metastases therapies are a priority for development with many recent introductions into the clinic of active treatments for a variety of tumor types. Systemic therapies aimed at the bone matrix (bisphosphonates (8,9) including zoledronic acid (10), denosumab, radiopharmaceuticals) and tumor cells (endocrine therapies, chemotherapy, molecular targets (11,12), and immunotherapy) are administered for disseminated disease. Local treatments (radiotherapy, surgery, high-intensity focused ultrasound, and cement augmentation) are used to control pain and to treat/prevent local complications (Fig. 1).

SKELETAL THERAPY ASSESSMENT TOOLS Accurate response evaluations of patients with bone metastases are notoriously difficult to do because measurable bony soft-tissue disease occurs infrequently. This is compounded by the fact that in breast and prostate cancer, bone-only disease is a common occurrence. The inability to accurately determine disease status in bone metastases results in the poor performance of conventional trial endpoints such as progression-free survival; the latter cannot be used as a surrogate of overall survival when disease is confined to bone (13). Thus, there is an overwhelming clinical and drug trial need to develop validated, noninvasive biomarkers to positively assess therapeutic benefit in patients with bone metastases (11,12,14). Symptom assessments (including pain scores and analgesic requirements) and development of skeletalrelated events are preferred markers of therapeutic efficacy in clinical trials of prostate cancer (15). More objectively, response can be gauged by a combination of imaging and clinical findings often used in combination with serum and urine biochemical markers of tumor burden and bone turnover (15,16). However, serum tumor markers of response are not available or are ineffective for the vast majority of tumors that metastasize to bone. For example, serum prostatespecific antigen (PSA) is not a reliable response biomarker in late stage, castrate-resistant prostate cancer (15). Short-term changes in PSA and CA15-3 by therapies targeting androgen/estrogen pathways may occur independently of effects on tumor cell growth or patient survival; temporary rises in these biomarkers (flare effects) are documented in patients who will eventually go on to benefit from therapies (15,17). Because of this, the prostate cancer working group 2 criteria (PCWG2) defines PSA progression criteria only, providing no response criterion (15). There are a number of serum and urinary markers of osteoblastic and osteoclastic

WB-DWI for Bone Metastases Monitoring

1051

Figure 1. Consequences of current and proposed benefits of new bone marrow assessments methods. Local and systemic therapies are applied for the treatment of metastatic bone disease. Current methods of assessing therapy response allow patients to be grouped using progression criteria. Functional imaging combined with “wet biomarkers” enable patient benefits to be positively identified, thus allowing more informed therapy choices.

activity that can potentially assess bone metastasis response to treatment (18), but by definition they only indirectly reflect disease activity within the metastatic bone marrow. Circulating tumor cells (CTCs) counts have emerged as powerful prognostic and response biomarkers for breast, colorectal, and prostate cancers (19–21), but cannot be recommended to guide therapeutic changes at this time (22).

Regardless of the method(s) used, current response biomarkers focus on assessing disease progression rather than positively addressing therapy benefit. Thus, patients are divided into progression/no progression categories rather than into the traditional groups of complete and partial response, stable, and progressive disease (Fig. 1). The categorization (progression / no progression) is explicitly acknowledged

Table 1 Working Protocol for Whole Body DWI Machine parameters (1.5T) Imaging sequence Repetition time Echo time No of slices Field of View Matrix Number of signal averages (NSA) Fat suppression scheme Diffusion encoding Parallel imaging factor Section thickness/gap b-values (s/mm2) Receiver bandwidth Acquisition time/station

Free breathing, single shot spin-echo echo planar imaging 9,000ms 68ms 200 (4 stations  50 slices/station) 450  450 cm (variable) 128  128 interpolated to 256  256 6 Short-tau inversion recovery (STIR, TI ¼ 180ms) Trapezoidal gradients, trace weighted, bipolar gradient scheme Grappa, pat factor 2, 30 reference lines 5 mm/0 mm overlap 50 and 900* 1628 Hz/pixel 5:51 minutes

b900 images from all four diffusion imaging stations are grouped and reconstructed as axial and coronal 5-mm slices. Whole body 3Dmaximum intensity projection of b900 images are displayed as rotating images (every 3 degrees ¼ 120 images) using an inverted grayscale. The ADC for all slices is calculated using the two b-values obtained monoexponential fitting of log signal intensity (SI) versus b values: S900 ¼ S50  exp ( b  ADC), where S900 is the signal intensity at b of 900 s/mm2 and S50 is the signal intensity at b of 50 s/mm2. *b50 s/mm2 images are used in preference to b0 values in order to partially mitigate the contribution of perfusion to calculated ADC values.

1052

Padhani et al.

Figure 2. Proposed biological model illustrating the relationships between bone marrow cell type and high b-value signal intensity and ADC values. Changes in signal intensity and water diffusivity caused by therapies are shown. The pattern of signal intensity and ADC change is dependent on the mechanism of therapy action and how rapidly repair processes begin.

within the PCWG2 criteria for the evaluation of metastatic castrate-resistant prostate cancer (15). Being unable to confidently identify therapeutic success using commonly available bone imaging methods such as bone scintigraphy and computed tomography (CT) scans means that clinical trials and care have focused on measures of progression as endpoints. PCWG2 recommends that time to progression or time to treatment failure be recorded as the primary outcome measures in clinical trials and definitive categorizations of response (complete/partial response and stable disease) should not be used (15). The clinical consequences of using progression criteria in patients with skeletal metastases include “prolonged exposure to potentially ineffective medications” (because there are no early clinically qualified biomarkers of therapy efficacy) and “all patients potentially getting all drugs — often too late” (combined effect of a lack of predictive theragnostic tests for the majority of tumors, ineffective early response biomarkers, and the short longevity of patients). Thus, currently available assessment methods can have negative impacts on oncologists’ thinking regarding therapy choices for patients with metastatic bone disease (Fig. 1). Imaging Methods Imaging methods have variable abilities to evaluate the bone mineral content, matrix including vascular-

ity or bone marrow. Technical limitations of data acquisition can limit the anatomic extent of coverage by a given imaging method. For example, dynamic contrast enhancement magnetic resonance imaging (DCE-MRI) cannot be used to evaluate the entire skeleton because of the need to perform a stationary series of rapid volume acquisitions before and after bolus intravenous contrast medium administration (23–25). Since no single imaging method can evaluate all aspects of metastatic bone disease, a comprehensive evaluation of the skeleton often requires a multimodal approach. 99mTc-MDP (technetium-99m-methylene diphosphonate) bone scans remain the commonest imaging method used to characterize and follow-up bone metastases. Unfortunately, bone scintigraphy and its higher spatial resolution positron emission tomography (PET) counterpart using 18F-sodium fluoride (18F-NaF) reflect only on the osseous component of bone. A positive bone scan occurs due to an osteoblastic response occurring secondary to an underlying bone abnormality and is thus an indirect indicator of metastatic bone marrow activity. False-positive bone scan lesions are a common occurrence due to fractures, arthritis, and benign bone lesions. Increased bone turnover resulting in the development of osteoblastic lesions is the reason for the high sensitivity of bone scans in prostate cancer. Recently, a semiquantitative estimate of skeletal MDP uptake (bone scan

WB-DWI for Bone Metastases Monitoring

1053

Figure 3. Bone marrow changes induced after G-CSF therapy and chemotherapy; a 50-year-old woman with metastatic breast cancer. First study (January 2012): End of capecitabine chemotherapy shows multiple bone metastases (arrows) on a background of reduced bone marrow signal intensity. Note the visibility of the entire length of the spinal cord and ready visibility of the dorsal nerve root ganglia in the cervical spine. There are bilateral, silicone containing breast enhancement bra pads in place which are not seen on subsequent examinations. Second study (February 2012): No interval treatment. Some disease progression and recovery of the background signal intensity of the bone marrow. Individual metastases are necrotic centrally. Third study (April 2012): Reassessment after 3 cycles of erubulin chemotherapy with 2 doses of growth-colony stimulating factor (G-CSF) given to prevent neutropenia in the second and third cycles of chemotherapy. G-CSF causes marked bone marrow hypercellularity and the resulting increases in signal intensity leads to the decreased visibility of the bone metastases. Note increased splenic size. Fourth study (May 2012): Rapid decreases in background bone marrow signal intensity after one further cycle of erubulin chemotherapy (no further G-CSF given). The bone metastases are easily visualized and appear larger. Note rising levels of the breast tumor marker CA-125, confirming nonresponse to therapy.

index) has been shown to provide prognostic information on metastatic prostate cancer (26,27) but its use as a response assessment tool has not been proven (26). The routine use of bone scans to guide early treatment changes is not recommended because scintigraphic/healing flare is recognized problem, occurring in 30% of patients usually within 3 months in patients responding to treatment (28,29). Drug trials using bone scans have criteria for progression (two categories only: no new lesions/new lesions) but not for response. To mitigate against healing flare reactions, apparent progression needs to be confirmed by follow-up bone scans when new focal “hot spots” have to be documented (15). However, patients with diffuse metastatic bony disease and bone superscans (where virtually all of the radiotracer is concentrated in the skeleton, with little or no activity in the soft tissues or urinary tract) cannot be followed for progression. Furthermore, the need to defer the decision of progression raises the issue of timeliness of the bone scan readouts for guiding clinical decision making. That is, patients have to be continued on potentially ineffective therapies before definite disease progression can be declared and therapy discontinuation/substitution implemented. In addition, bone scintigraphy is unsuited for the therapy assessment of predominantly lytic disease without an asso-

ciated osteoblastic reaction as typically seen in renal or thyroid cancers (that is, cold spots on bone scans cannot be assessed for therapy benefit). CT scans are also limited in their ability to assess therapy response because bone structure rarely normalizes even with completely effective therapy. The appearance of new or worsening bone sclerosis on CT in patients is occasionally erroneously classified as disease progression (CT flare response) by the inexperienced radiologist (30). RECIST (v. 1.1) criteria do allow individual osteolytic or mixed osteolytic/osteoblastic metastases to be measured if there is a soft tissue component, but diffuse disease and osteoblastic bone metastases are considered nonmeasurable (31,32). The MD Anderson (MDA) Cancer Center criteria recognize bony sclerosis as a response criterion when there are other signs of response and in the absence of progressive bone disease (31,33). The MDA criteria, which are not widely used, do not incorporate measurements of CT density and patients receiving commonly administered anti-osteoclastic drugs (bisphosphonates including zoledronic acid and denosumab) are ineligible for the CT assessment of response. A number of PET tracers have been evaluated for their ability to predict and monitor bony therapy response due to high tumor to background uptake

1054

Padhani et al.

Figure 4. Poor visibility of treated metastases and osteoblastic metastases. A 69-year-old with metastatic prostate cancer on long-term, third-line hormonal therapy with abiraterone being evaluated for rising serum PSA levels. He has had an excellent response to 2 years of treatment with residual abnormalities in his bone marrow visible on T1W (first left image) and T2W (with fat suppression) spinal images. No hyperintensity is seen on the b900 3D MIP (inverted scale) image (third image), indicating the absence of osteolytic disease. Bone scan (right side image) shows a focal area of osteoblastic uptake in the intertrochanteric region of the left femur (arrow) which is not visible as a discrete region on the b900 3D inverted MIP image.

ratio of radiopharmaceuticals (34–37). Nontumor type-specific tracers include 18F-flurodeoxyglucose (FDG), 11C/18F-thymidine (FLT), 11C/18F-acetate, and 11C/18F-choline (38). 18F-FES (fluoroestradiol for breast cancer), 18F-FDHT (fluorodihydrotestosterone for prostate cancer), and engineered antibody fragments that targeting tumor cell-surface targets such as HER-2/neu (for breast cancer) are examples of tumor specific tracers (39). Most studies have focused on FDG-PET, which is discussed in detail below. Readers are invited to review the literature quoted above for more in-depth discussions on the potential utility of other more rarely used radiotracers. Studies have shown that lytic bony disease is usually FDG-avid and osteoblastic lesions are less FDGavid (40,41). As a result, FDG-PET scans perform better in patients with breast cancer than in prostate cancer. FDG-PET scans tend to be positive in highergrade disease and baseline specific uptake values (SUV) do correlate with prognosis in some tumor types. Several FDG-PET scan response criteria have been described, with the PERCIST criteria gaining widespread acceptance (42). These criteria recognize that residual FDG uptake can be seen despite effective therapy, possibly due to macrophage activity, and therefore complete metabolic response is defined as a decrease in tumor SUV to the level of surrounding normal tissue (not the complete absence of FDG uptake). New foci of FDG activity in the absence of

anatomic evidence of disease progression still need to be verified on follow-up scans before therapy progression is declared. Tumor size measurements remains important under PERCIST, with increases or decreases in size of lesions without metabolic changes requiring clarification by follow-up scans, when further changes in size or metabolic uptake or new disease is needed to declare progression. Importantly, FDG-PET response criteria do allow the categorization of patients into traditional response grouping, thus enabling the positive identification of therapy benefit.

MRI for Assessing Bone Metastases A number of MRI methods can evaluate bone for metastasis detection and response assessments (43). Relevant sequences include T1-weighted spin-echo, T2-weighted (with or without fat suppression), and short tau inversion recovery (STIR) which are sensitive to the cellular, fat, and water content of the bone marrow (44). Gradient-echo proton density/T1 sequences using two or three point Dixon techniques (yielding in-phase, opposed phase, water-only, fat-only, T2*, and water and fat fraction images) can be used to objectively evaluate the relative non-fat and fat content of bone marrow (45). Susceptibility-weighted (T2*) sequences are sensitive to spin dephasing induced by trabecular bone (44). Ultrashort TE (UTE) sequences can also evaluate trabecular bone

WB-DWI for Bone Metastases Monitoring

1055

Figure 5. Proposed scheme for assessing therapy response of bone metastases using diffusion-weighted MRI scans, ADC measurements, and morphologic images. Squared boxes are instructional. Observations are indicated within diamond boxes. Possible biologic inferences/explanations are indicated within open boxes.

structure in healthy bone and metastatic disease (46). A number of studies have evaluated bone marrow vascularization using DCE-MRI techniques (23–25). Diffusion-weighted imaging (DWI) is increasingly being used because of its sensitivity to bone marrow cellularity, the relative proportions of fat and marrow cells, water content, and bone marrow perfusion (25,47). Since multiple sequences can be performed in any patient examination, only MR can be considered a truly multifunctional single (nonhybrid) evaluation method for evaluating bony metastatic disease. Whole-body DW imaging (WB-DWI) is emerging as a promising bone marrow assessment tool for detection and therapy monitoring of bone metastases (48–50). Advantages of WB-DWI include the fact that no ionizing radiation is administered and no injection of isotopes or any contrast medium is necessary. Importantly, whole-body skeletal examinations are possible in reasonably short data acquisition times. Disease depicted on high b-value images can be volumetrically quantified and tissue water diffusivity maps (apparent diffusion coefficient (ADC); units 

103 mm2/s or mm2/s) can enable spatial heterogeneity of tissues/tumors to be analyzed, before and in response to treatment. ADC values, which are theoretically independent of magnetic field strength, and the relative simplicity of data acquisition can facilitate multicenter and longitudinal studies being undertaken (51). However, comparison of DW-MRI data acquired between different equipment manufacturers may not always be straightforward (52,53). One major issue concerns WB-DWI data acquisition at 3.0T. WBDWI is currently best performed at 1.5T because of the ability to uniformly suppress fat over large fields of view. Although WB-DWI at 3.0T has the potential to improve image signal-to-noise ratios (SNRs), it is more technologically challenging to do so due to the greater incidence of susceptibility artifacts and poorer fat suppression over large fields of view. Recent technological advances including improved shimming routines and multitransmit equipped MRI systems do allow WB-DWI data acquisitions at 3T (53,54). We prefer performing WB-DWI at 1.5T and our protocol is detailed in Table 1. In addition to the diffusion

Figure 6. Disease progression in breast cancer. Serial changes in a 71-year-old woman with progressive metastatic breast cancer being treated with exemestane (aromatase inhibitor) and bisphosphonates. a: b900 3D MIP (inverted scale). A diffuse pattern of metastatic bone disease is seen progressively increasing in extent and signal intensity indicating disease progression. Whole-spine sagittal T1-weighted images (not shown) demonstrated disease progression between the first and third examinations with no further appreciable changes between the third and fourth examinations. T2-weighted sequences with spectral fat suppression (not shown) did not change appreciably between examinations. b: Spread plots showing ADC changes with corresponding calculated descriptive histogram statistics of a pelvic volume of interest defined on examination 4 (using b900 images) and retrospectively applied to all examinations after robust, elastic image registration. A unimodal, nonnormal histogram with a positive skewness and kurtosis is observed on all four studies. ADC values decrease over time without changes in standard deviation consistent with increasing cellularity. The majority of pixels are less than the 1500 mm2/s (vertical control line) for all examinations. c: Left column: pixel scatterplots of muscle normalized b900 signal intensity (x-axis) and ADC values (y-axis). Middle and right columns: thresholded muscle normalized b900 and ADC histograms showing serial changes over time (superior to inferior). The control lines on the histograms and scatterplots are placed on 3 and 9 for normalized b900 signal intensity and 650 and 1500 mm2/s for ADC. The images show that ADC values decrease (lower mean ADC values on the ADC histograms) and signal intensity values increase (more green pixels) with tumor progression. Corresponding shifts in the SI-ADC scatterplots are visible with a movement of the scatterplot to the right side over time. The changes in the scatterplot are in the opposite direction to that observed with successful therapy response (see Figs. 8, 9). Analyses were done using Oncotreat software (Siemens HealthCare, works-in-progress). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

WB-DWI for Bone Metastases Monitoring

1057

Figure 6. (Continued)

sequences, we also acquire 1) whole spine: T1-weighted, turbo spin-echo sagittal images (acquisition time 2:21 minutes); 2) whole spine: T2-weighted, turbo spin-echo sagittal images with spectral fat suppression (acquisition time 2:36 minutes); 3) whole-body (vertex to upper mid thighs): T1-weighted, gradient-echo axial 2-point Dixon sequence (acquisition time 3:00 minutes) that automatically generates four image-sets (in-phase, opposed phase, water-only [WO], and fat-only [FO]) from which image fat% and image non-fat% images can be calculated. Finally, whole-body (vertex to upper mid thighs): T2-weighted, STIR axial images with half-Fourier single shot turbo spin-echo (HASTE) readouts (acquisition time 4:00 minutes) is also undertaken. Detailed scanning parameters for each sequence have been published (47,53); the entire examination takes 52 minutes to complete. Using this protocol, our combined institutions have performed more than 2300 patient examinations over the last 5 years; 60% in breast cancer, 10% in myeloma, 15% in prostate cancer, 9% in renal cancer, 3% Melanoma, and 3% for other malignancies.

NORMAL BONE MARROW IMAGING WITH WB-DWI Visual inspection of WB-DWI is excellent at demonstrating the variability of the normal bone marrow dis-

tribution and changes induced by therapeutic manipulations. The normal adult bone marrow distribution becomes established by 25 years of age; the red bone marrow is found predominantly in the axial skeleton, whereas yellow bone marrow predominates in the peripheral skeleton; this pattern is readily visible on WB-DWI. There is variable red bone marrow atrophy and trabecular bone loss after 40 years of age (55), particularly in women (possibly related to estrogen deficiency and osteoporosis (56)), resulting in increased relative adiposity and thus lowering signal intensity of axial bone marrow. Regional variations in the bone marrow cellularity are reflected in the signal intensity of high b-value DWI. Because yellow bone marrow has a lower water and cellular content (10%– 20%) (57–60) its signal intensity on DWI is low. On the other hand, with increasing cellularity (59,61) and water content (40%–60%), mixed yellow-red bone marrow returns higher signal intensities. Unlike the inverse correlations between ADC and cell density seen in many soft-tissue tumors (62–68), ADC changes with increasing bone marrow cellularity are nonlinear (69–73). This finding can be related to the distribution of cell sizes within the confines of the bone marrow space. Thus, the relative number of larger fat and smaller myeloid/erythropoietic/tumor cells that reside within the bone marrow affects the

1058

Padhani et al.

Figure 7. T2 shine-through-like pattern indicating successful response to therapy in a 42-year-old female with metastatic breast cancer. Treated with continuous capecitabine chemotherapy and bisphosphonates. a: Serial b900 3D MIP (inverted scale) images. Widespread multifocal bony metastatic disease within the vertebral column, ribs, pelvis, and proximal femora. Signal intensity alterations of individual lesions occur very slowly despite an excellent clinical response to treatment. b: Whole-spine sagittal T1-weighted images. Metastatic lesions appear to have sharper margins and have increasing fat within metastatic lesions (arrows) over time, consistent with a healing response to therapy. c: Whole-spine sagittal T2-weighted sequences with spectral fat suppression. An increase in signal intensity on T2W within lesions involving the upper dorsal vertebrae in keeping with an increase in water content. Note the dark signal intensity within lesions due to fat suppression at later timepoints. d: Spread plots showing diffusivity changes for a pelvic volume of interest defined on examination 1 (using b900 images) and applied to all examination after robust image registration. A marked increase in ADC values occurs after starting therapy that does not change appreciably over time consistent with decreased cellularity. The vertical control line is placed at 1500 mm2/s.

appearances of the bone marrow on DWI. Yellow bone marrow has lower signal intensity and low ADC values (69,73,74), probably because of the reduced proton density, the hydrophobic nature of fat, and lower bone marrow perfusion (compared to red bone marrow) (75). With increasing bone marrow cellularity (which displaces large fat cells and increases the vascularity of the bone marrow), the signal intensity on DWI increases, but appears to paradoxically return higher ADC values compared to yellow bone marrow (71–74,76–78). However, once all bone marrow fat cells are displaced, further increases in bone marrow

cell density within the confines of a fixed marrow space may cause ADC values to plateau or even decrease, although the extent of the latter effect has not yet been comprehensively documented (Fig. 2). Common causes of bone marrow hypointensity in cancer patients include chemotherapy, radiation treatment, and drugs that cause osteopenia. Osteoporosis-related relative increases in bone marrow adiposity (70,79,80) have characteristic appearances on high WB-DWI such as the absence of the normal high signal intensity of the bone marrow and the resulting visibility of the entire length of the spinal cord and cervical and lumbar nerve

WB-DWI for Bone Metastases Monitoring

1059

Figure 7. (Continued)

roots ganglia (Fig. 3). A potential area of difficulty when assessing bone metastases is the effect of bone marrow hyperplasia on the background bone marrow signal. Growth colony stimulating factor (G-CSF) usage for example causes marked increases in the bone signal intensity due to increases in bone marrow cellularity and water. In this case, bony metastases can become less conspicuous against increasing background signal intensities (Fig. 3). On the other hand, the detection of bone metastases is improved in bone marrow that is relatively hypocellular, for example, in the older patient.

SKELETAL METASTASES DETECTION On WB-DWI, lytic/infiltrative skeletal metastases appear as focal or diffuse areas of high-signal inten-

sity on high b-values on a background of lower signal intensity of the normal bone marrow. It is important to emphasize that metastasis detection on WB-DWI should not be done in isolation but rather has to be considered as a valued adjunct to anatomical wholebody MRI assessments (50). This assertion has been highlighted in a recent meta-analysis that demonstrated that the high sensitivity of WB-DWI to detect bone metastases was at the expense of specificity (48). Thus, the pooled sensitivity/specificity of wholebody MRI (with DWI) has been reported as 87.7% (95% confidence interval [CI]: 76.3–94.9%) and 86.1% (95% CI: 79.2–91.4%) compared to 90.9% (95% CI: 84.3–95.4%) and 96.1% (95% CI: 92.2–98.4%) for whole-body MRI without DWI (48). Causes for false-positive findings on WB-DWI include bone marrow edema caused by fractures,

Figure 8. Therapy response in metastatic breast cancer. Serial changes in a 41-year-old woman with responding metastatic breast cancer being treated with capecitabine chemotherapy and zoledronic acid. a: b900 3D MIP (inverted scale) images. A diffuse pattern of metastatic bone disease is seen to progressively decrease in extent and signal intensity indicating disease response. Small volume liver disease can be seen on the first examination (arrowhead). On the second image, the bone lesions are seen to have decreased in their signal intensity and appear more “hazy.” On the third examination two new liver metastases were seen (arrows) which underwent radiofrequency ablation. On the fourth examination, another liver lesion (arrow) appears which was also ablated. There are a few new focal areas of increased signal intensity on the fifth examination (arrows) indicating possibly relapse of bone disease. b: Whole-spine sagittal T2-weighted with spectral fat suppression. Dates are identical to (a). Images show increased signal intensity of the bone marrow on examination 2, indicating increasing bone marrow edema. Loss of vertebral height at D12 also. By examination five, the bone marrow signal intensity has decreased, indicating loss of tissue water. c: Whole-spine sagittal T1-weighted images showing decreases in signal intensity on examination two due to bone marrow edema—this appearance can mimic disease progression (sometimes called T1-pseudoprogression). Dates are identical to (a). d: Spread plots and cumulative frequency histograms showing ADC changes with corresponding descriptive statistics of a pelvic VOI (using b900 images) defined on examination 1 and applied to all examinations after robust image registration. A unimodal, nonnormal histogram with a positive skewness and kurtosis is observed initially. Mean ADC values increase initially with a negative skewness and reduced kurtosis consistent with decreasing cellularity. The majority of pixels are greater than 1500 mm2/s (vertical control line) for the second and third examinations. ADC values decrease over time with an increasing spread of ADC values in examinations 4 and 5. Cumulative histograms are an alternative way of representing the same data. e: Left column: pixel scatterplots of muscle normalized b900 signal intensity (x-axis) and ADC values (y-axis). Middle and right columns: thresholded muscle normalized b900 and ADC histograms showing serial changes over time (superior to inferior). The control lines on the histograms and scatterplots are placed on 3 and 9 for normalized b900 signal intensity and 650 and 1500 mm2/s for ADC. A pelvic VOI defined on examination 1 (using b900 images) was applied to all examinations after robust image registration. The images show initial increases in ADC values accompanied by reductions in signal intensity. Late reductions in ADC values and b900 signal intensity can be ascribed to the onset of bone marrow repair mechanisms, including removal of dead tumor cells, loss of tissue water, bone sclerosis, fat deposition, and reduced perfusion. Contrast these patterns of changes in tumor progression shown in Fig. 6c. Analyses were done using Oncotreat software (Siemens HealthCare, works-in-progress).

WB-DWI for Bone Metastases Monitoring

1061

Figure 8. (Continued)

osteoarthritis, infection, bone infarcts, vertebral hemangiomas, isolated bone marrow islands, and bone marrow hyperplasia due to G-CSF. Many of these false-positive findings can be overcome by corre-

lating high b-value DW images with corresponding ADC maps and anatomical T1W/T2W images. In contrast, causes for false-negative findings include low levels of bone marrow infiltration such as

1062

Padhani et al.

Figure 8. (Continued)

in smoldering multiple myeloma or when background bone marrow hyperplasia obscures the presence of metastases (Fig. 3). Similarly, the detection of skeletal metastases on WB-DWI may be impaired in areas of body movement such as the ribs and occasionally in the sternum. Visibility of skull vault infiltrations can also be impaired because of the adjacent high signal of the brain. The visibility of skull base disease is impaired because of susceptibility effects. Another important cause for false-negative findings is successfully treated malignant disease and sclerotic metastases (Fig. 4). In general, lytic/infiltrative lesions are better detected than sclerotic or treated lesions on WB-DWI. This is due to the lower water and cellular content of sclerotic and treated metastases (81). DWMRI is better at detecting skeletal lesions from smaller cancer cell infiltrations such as those due to breast cancer, myeloma, lymphoma, small cell cancers, and neuroendocrine tumors compared to bony metastases from larger clear cell cancers of the kidneys; the latter are sometimes relatively less conspicuous. From a review of the prostate literature, it is clear that WB-DWI detects more malignant lesions per patient and fewer benign lesions (fewer false positives) compared to planar bone scans (82). However, WBDWI has lower sensitivity than 18F-NaF PET/CT but has higher specificity (83) because false positives occur with less frequency, particularly when anatomic MRI correlations are made. Concerning the evaluation of men at high risk for developing metastases from prostate cancer, a recent study showed that WB-DWI

outperforms bone scans with targeted x-rays in detecting bone metastases and performs as well as CT for lymph node evaluation (50), enabling WB-DWI to become the tool of choice for this indication.

BONY METASTASES THERAPY MONITORING ON WB-DWI When multiple WB-DWI studies are compared across time, it is important to normalize the signal intensity of the high b-value thick volume, maximum intensity projections (MIP) images for effective comparisons to be made. Normalization between the different imaging stations at a single timepoint and between different timepoints needs to be undertaken; this is an area of active research currently and optimal methods have yet to be established. In practice, therefore, between different timepoints normalization is done by setting the window level to that of a tissue assumed to be unchanging between examinations, then maintaining the window width between examinations. We often use the kidney or brain signal as normalization tissues, recognizing that impaired renal function can alter diffusionweighted signal intensities of the kidneys. Therapy assessments on WB-DWI are made by observing changes in the volume and symmetry of signal intensity abnormalities on high b-value images together with changes in ADC values. Crosscorrelating DWI findings with morphological appearances on T1W, fat-saturated T2W/STIR, and Dixon

WB-DWI for Bone Metastases Monitoring

1063

Figure 9. Response with small ADC increases. Serial changes in a 53-year-old woman with metastatic breast cancer being treated with aromatase inhibitors and bisphosphonates. a: b900 3D MIP (inverted scale) images. A diffuse pattern of metastatic bone disease that progressively increases in the extent and signal intensity indicating likely disease response. Note the disappearance of a right sided pelvic lymph node (arrow). Whole-spine T1-weighted and T2-weighted images with spectral fat suppression images showed diffuse bone marrow infiltration showing no discernable changes over time (not shown). b: Left column: pixel scatterplots of muscle normalized b900 signal intensity (x-axis) and ADC values (y-axis). Middle and right columns: thresholded muscle normalized b900 and ADC histograms showing serial changes over time (superior to inferior). The control lines on the histograms and scatterplots are placed on 3 and 9 for normalized b900 signal intensity and 650 and 1500 mm2/s for ADC. A pelvic VOI defined on examination 1 (using b900 images) was applied to all examinations after robust image registration. An increase in ADC values is seen but the majority of pixels stay below the 1500 mm2/s vertical control line. This is in contrast to more marked ADC increases observed when bony metastases are successfully treated with chemotherapy (Figs. 7d, 8d). By examination 3, a marked decrease in ADC values accompanied by an increase in variance can be seen. c: Columns represent b900 diffusion-weighted images (left column), image fat (middle), and image water/non-fat fraction (right column). The latter two were derived using a T1-weighted 3D 2-point Dixon sequence which allows the generation of in-phase and opposed phase images as well as water-only and fat-only images. The water- and fat-only images are used to calculate image fat and non-fat fractions. Decreases in diffusion-weighted signal intensity are accompanied by decreases in image water fraction and corresponding increases in image fat fraction (vertical arrows). d: Histogram analysis of changes between examinations 1 (December 2011) and 3 (July 2012). Top row shows colored ADC maps (scale 0–1500 mm2/s) of the outlined pelvic bone marrow VOI defined using b900 images. The left top image represents the baseline examination and the top right image the posttreatment examination. The second row shows the raw data histograms for the two examinations (pretreatment: blue; posttreatment: orange). The forth row shows threshold histograms (threshold values used were 650 and 1500 mm2/ sec) with corresponding proportions of pixels within the data ranges. The corresponding thresholded ADC maps are shown in row 3. Row 5 shows a pixel difference histogram using a cutoff value of 200 mm2/sec with corresponding ADC difference pixel map (bottom left image). Red pixels have increased their ADC values by 200 mm2/sec and blue pixels have reduced their values by 200 mm2/sec. Analyses were done using Oncotreat software (Siemens HealthCare, works-in-progress).

images is important. Lesion-by-lesion signal intensity and ADC value changes can be interpreted using the guidance in Fig. 5 with several distinct patterns being recognized in the therapy assessment setting. 1. Increases in the volume of previously documented abnormal signal intensity, new areas of abnormal signal intensity, or increases in the intensity of abnormalities on high b-value DW images can indicate disease progression (Fig. 6). Modest increases, unchanged, or slight decreases

in ADC values compared to pretherapy values can occur (84,85) in the setting of progression. Reductions in ADC values are probably related to increasing cellularity within a fixed bone marrow space (Fig. 6). Stable ADC values could occur with unchanged tumor cellularity accompanied by increases in the geographic extent of disease. The causes for modest increases in ADC values with disease progression are identical to those discussed above on bone marrow cellularity. To briefly reiterate, increasing tumor infiltration

1064

Padhani et al.

Figure 9. (Continued)

displaces fat cells, increases bone marrow water (including water in the extracellular space), and increases tissue perfusion, thus returning higher ADC values compared to yellow or mixed bone marrow (71–74,76–78). The important point to

note is that increases in ADC values with progression tend to be of small magnitude, provided that the metastatic lesions remain nonnecrotic (85). In our experience ADC increases >1400–1500 mm2/s are rarely seen with disease progression unless

WB-DWI for Bone Metastases Monitoring

1065

Figure 9. (Continued)

there is de novo tumor necrosis. In a recent analysis only 5% of tumor regions of interest in 71 patients with untreated/relapsed malignant lesions (breast, prostate and renal cancers, myeloma and melanoma) had ADC values 1500 mm2/s (for an imaging protocol where b50–b900 s/ mm2 diffusion gradients were used for data acqui-

sition) (internal data). Readers should be able to determine their own cutoff values for their imaging acquisition protocol by evaluating the range of ADC values of untreated disease lesions (73). 2. Occasionally, unchanged high signal intensity on high b-value images associated with marked rises in ADC values is observed (Fig. 7). Although

1066

Padhani et al.

Figure 10. Responding disease with no changes in ADC values. A 60-year-old female with metastatic breast cancer on capecitabine and bisphosphonates therapy. Clinical follow-up and CT imaging after 2.5 years shows a sustained response to therapy. a: b900 3D MIP (inverted scale) images. Reductions in signal intensity in the majority of lesions within the vertebral column and pelvis suggesting response to treatment. b: Whole-spine sagittal T1-weighted images. Individual metastatic lesions become better defined and appear darker consistent with a sclerotic response to therapy. Dates are identical to (a). c: Whole-spine sagittal T2-weighted sequences with spectral fat saturation. Reduced signal intensity of the bony metastases within upper dorsal spine and in the lumbar spine implies reduced water-content. Dates are identical to (a). d: Left column: pixel scatterplots of muscle normalized b900 signal intensity (x-axis) and ADC values (y-axis). Middle and right columns: thresholded muscle normalized b900 and ADC histograms showing serial changes over time (superior to inferior). The control lines on the histograms and scatterplots are placed on 3 and 9 for normalized b900 signal intensity and 650 and 1500 mm2/ s for ADC. A pelvic VOI defined on examination 1 (using b900 images) was applied to all examinations after robust image registration. No appreciable change in ADC values is seen, with the majority of pixels staying below the 1500 mm2/s vertical control line. e: CT scans showing increases bony sclerosis in lytic/infiltrative disease areas seen on baseline scans (arrows). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

WB-DWI for Bone Metastases Monitoring

Figure 10. (Continued)

1067

1068

Figure 10. (Continued)

this combination of observations is commonly termed “T2 shine-through,” readers should note that changes in tissue T1 relaxation times can also lead to these appearances, particularly on inversion recovery diffusion-weighted sequences. Regardless, this pattern indicates that there has been a successful response to therapy (Fig. 3). This pattern is commonly seen in patients with multiple myeloma, lymphoma, and occasionally in other solid metastatic neoplasms that respond successfully to therapy. Lesions showing this pattern can take a long time (often more than 1 year) for high b-value signal intensity to decrease to the background levels, as discussed below. 3. Decreases in bone marrow disease signal intensity on high b-value images are generally observed with successful treatments. Effective tumor cell death should result in greater water diffusivity manifested as higher ADC values (84,86). The extent of ADC increases seems to depend on the type of treatment given (Fig. 3). We have noted that ADC increases are greater for cytotoxic chemotherapy and radiation (Fig. 8). The latter treatments cause tumor cell death via a number of mechanisms (apoptosis, necrosis, mitotic catastrophe, autophagy, and senescence), many of which lead to tumor necrosis with an

Padhani et al.

inflammatory component (87–91). Increases in extracellular bone marrow water can be seen on corresponding T1-weighted and fat-suppressed T2-weighted/STIR images with focal or diffuse signal intensity changes. On T1-sequences, such bone marrow edema can superficially mimic disease progression (so called T1-pseudoprogression)—an important pitfall for the unwary (Fig. 8). On fat-suppressed T2-weighted/STIR images high tissue water is observed as signal intensity increases. When patients are treated successfully with hormonal therapies, ADC value increases seem to be less marked than chemotherapy or radiation possibly because cell death is less likely to be associated with inflammation (87–91) (Fig. 9). Corresponding, T1-weighted and fatsuppressed T2-weighted/STIR images may not show pronounced signal intensity changes, probably because bone marrow tissue water appears less affected. 4. Occasionally high b-value signal intensity decreases are associated with unchanging or slight ADC values decreases (ie, no ADC increases). Although biophysical mechanisms are not completely clear, by observing changes on morphologic images, Dixon water, and fat images and on CT scans, we have ascribed this observation to the development of bone sclerosis, myelofibrosis, and to the return of fat to the bone marrow. Generally, this pattern generally occurs in clinical responders (Fig. 10), although very occasionally we have noted it in nonresponding patients (so-called sclerotic progression). Since this pattern can be seen in responders and occasionally in nonresponders, these appearances should be considered as indeterminate and currently we resort to morphologic and clinical assessments to assign the final response category. 5. Stable disease despite therapy can be identified by noting unchanging appearances on high bvalue images. ADC changes can be variable, often remaining stable but are sometimes slightly decreased, presumably because of increases in cell density within lesions that are unchanging in their extent (85). Slight increases in ADC values can also be observed generally staying well below the threshold required for response (Fig. 5). As already pointed out, the above patterns become evidence relatively soon (within 1–3 months) after instituting therapy. However, the onset of changes for the different response categories are yet to be firmly defined. The long-term changes observed on WB-DWI with repair and remodeling of the bone marrow are also not yet comprehensively described. We have noted that bone marrow metastatic disease responding successfully to therapy ultimately reduces in signal intensity on high b-value images accompanied by reductions in ADC values (Figs. 8–10). This probably occurs via a number of mechanisms including removal of dead tumor cells, the development of bone sclerosis, reemergence of bone marrow fat, loss of tissue water, secondary myelofibrosis, and decreased tissue perfusion (92)

WB-DWI for Bone Metastases Monitoring

1069

Figure 11. Value of histogram analysis to supplement visual analysis of WB-DWI. A 42-year-old woman with breast cancer who received sequential chemotherapy with bevacizumab and bisphosphonates for metastatic bone and liver disease. There was an initial response then disease progression on chemotherapy. Between January and June 2011 she received paclitaxel chemotherapy and between August 2011 and April 2012 she received capecitabine. a: Zoomed b900 3D MIP (inverted scale) images. Liver metastases are visible. Decreases in the signal intensity of bone marrow metastases to background levels results in the apparent disappearance of the metastatic lesions. The background signal intensity also increases (no G-CSF was administered). Metastases in the bone marrow appear to reemerge by June 2011. The focal hyperintensity in the uterus represents a uterine polyp. Between August 2011 and April 2012 therapy response to capecitabine is difficult to judge using visual assessments alone. b: Zoomed T1-weighted spinal images show the deposition of intratumoral fat within the L1 lesion by April 2011 (arrow). The deposited fat disappears with tumor relapse, although there is a transient increase in intratumoral fat on the December 2011 and March 2012 examinations. In general, the T1-weighted images show disease progression after switching to capecitabine chemotherapy. Dates are identical to (a). c: Zoomed T2-weighted spinal images with spectral fat suppression show a transient increase in signal intensity within L1 when treated with paclitaxel. No increases in signal intensity are seen after switching to capecitabine chemotherapy. Dates are identical to (a). d: Spread plots showing ADC changes of pelvic VOIs (using b900 images). The VOIs defined on examinations 1 and 8 are shown (top and bottom right images; 374.8 cm3 and 361.9 cm3, respectively). Note that the VOIs involve different anatomic areas with a degree of overlap. VOI 1 was applied to examinations 1–3. VOI 1 and VOI 8 were combined and then applied to examinations 4–8 after robust image registration. There is a substantial initial increase in ADC values on examination 2, with the majority of pixels above the 1500 mm2/s vertical control line, indicating responding disease. By examination 3, a bimodal distributions develops and on the relapsed scan (examination 4), voxel values are lower than pretherapy values. Changing to capecitabine chemotherapy results in minor initial ADC increases which then fall as the patient progresses. e: Serial changes in relative frequency histograms with corresponding descriptive statistics of the data shown in (d). The number of x-axis bins and y-axis scale has been fixed to more easily appreciate changes over time. A unimodal, nonnormal histogram with a positive skewness and kurtosis is observed initially. Mean ADC values increase initially on examination 2 with a negative skewness but no changes in kurtosis consistent with decreasing cellularity. The development of a bimodal histogram on examination 3 results in decreasing kurtosis. By the time of relapse (examination 4) positive skewness and kurtosis have reemerged. The visualized change in histogram kurtosis between examinations 4 and 5 is not captured in the kurtosis values due to development of “fat tails” on histogram 5, which can elevate kurtosis values. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

1070

Padhani et al.

Figure 11. (Continued)

(Fig. 3). These water diffusivity changes seem to occur slowly generally becoming visible many months after starting therapy (93,94), depending on the tumor type,

type of therapy administered, and initial therapy response. We have already noted that for the “T2 shinethrough” pattern the time course for signal intensity

WB-DWI for Bone Metastases Monitoring

1071

Figure 11. (Continued)

and ADC change to decrease can be very prolonged (more than 1 year) (Fig. 7). Of course, if tumor relapses within the bone marrow, then the signal intensity on high b-value images and ADC changes develop corresponding progression appearances, as described above. Thus, in the clinical setting where multiple local and systemic therapies have been administered, there will be active disease intermixed with successfully treated areas, resulting in a heterogeneous signal intensity distribution and ADC values (Fig. 11). Response assessments by visual appearances can therefore become problematic for multiply treated patients and quantitative methods of

evaluating diffusivity changes including histogram analyses can be helpful for these situations.

QUANTITATIVE WB-DWI ANALYSIS Volume of Interest Definitions Quantitative tumor volume assessments are usually undertaken by segmenting high signal intensity regions on high b-value images. Such high signal intensity regions will include “cellular dense viable tumor” recognizing that not all variable tumor will be depicted. We have already noted that low levels of

1072

Padhani et al.

Figure 12. Volume image registration and volume of interest segmentation. Serial scans in a 62-year-old man with metastatic castrate-resistant prostate cancer with progressive disease following prolonged first-line androgen deprivation for metastatic bone disease. He initially received radiotherapy to the bony pelvis and left femur for symptomatic disease and then docetaxel chemotherapy for further disease progression. a: b900 3D MIP (inverted scale) images. Examination 1 show multifocal metastases with edema around the proximal part of the left femur. Field placements for the right hemipelvis and left femur radiotherapy are indicated. On the first follow-up scans there is disease progression in the spine and ribs. The irradiated tumor regions also look larger. There are marked improvement in appearances after 4 (examination 3) and 10 cycles of docetaxel (examination 4) with corresponding reductions of serum PSA levels (normal 1400– 1500 mm2/s. When tumors respond successfully to therapy, kurtosis values generally decrease and the standard deviation/variance increases (Figs. 7d, 8d, 9d). Negative skewness (tail to the left) often develops in responding disease if the histogram retains a unimodal shape (Fig. 8d). Thus, the transformation of a

Pre-therapy plus post therapy masks Pre-therapy mask “Lesion-over-time” used for progressive & mixed response lesions

*ADC ranges for normal, viable tumor, and nonviable tumor are likely to be tumor type and acquisition protocol dependent. Their values can be determined by evaluating normal subjects and untreated patients using the imaging protocol of choice.

Not applicable as VOIs are of different sizes ditto ditto

Used to show location & and extent of ADC change at the voxel level Used to show location & proportion of voxels in normal, viable & non-viable ranges* Pre-therapy mask

Pre-therapy mask

Estimated from VOI segmented on high b-value images & the proportion of viable voxels determined on threshold histograms* ditto

Used to document ADC changes using measures of central tendency, variance measures, skewness & kurtosis

1075

“Region-over-time” used for responding lesions

Frequency histogram & map Post-therapy study Pre-therapy study Type of analysis according to therapy response

Volume of interest recommendations

Table 2 VOI Definitions and Type of Histogram Analysis

Viable tumor volume

ADC histogram tools

Threshold histogram & map

Difference histogram & map

WB-DWI for Bone Metastases Monitoring

pretreatment kurtotic, positively skewed unimodal ADC distribution into a stretched complex shape in response to a therapy intervention with a substantial number of pixels with ADC values