Dual energy X ray absorptiometry (DEXA) is an X ray imaging technique primarily used to derive the mass of one material in the presence of another through knowledge of their unique X ray attenuation at different energies. DXA is an extremely accurate and precise method for quantifying bone mineral density (BMD) and mass body composition assessment. DEXA is one of the most accurate and precise methods for quantifying BMD and mass in vivo. Bone ‘mineral mass, primarily consisting of hydroxyapatite, is the mineral component of bone that is left after a bone is defleshed, lipids extracted and ashed. The nature of the DXA system is that it creates a planar (two dimensional) image that is the combination of low and high energy attenuations. Although density is typically thought of as a mass per unit volume, DXA can only quantify the bone density as a mass per unit area, since it uses planar images and cannot measure the bone depth. DEXA defines the composition of the body as three materials having specific X ray attenuation properties: bone mineral, lipid (triglycerides, phospholipid membranes, etc.) and lipid free soft tissue. The non-lipid soft tissue mass (STM) is the sum of body water, protein, glycerol and soft tissue mineral mass. For each pixel in a DXA image, these three mass components are quantified. MEASUREMENTS FROM DXA SCANS There are relatively few values reported from DXA body composition systems. They are listed here. The bone measures are available from all DXA scan modes while the body composition measures are only available from the whole body scan mode. Bone mineral content (BMC). BMC is the mineral mass component of bone in the form of hydroxyapatite, Ca10(P04)6(OH)2. BMC is typically measured in grams. Note that BMC does not include the mass of any of the organic components of bone (marrow, collagen, etc.). Thus, accuracy can only be assessed against ashed bone samples. Bone area (BA). BA is the projected area of the bone onto the image plane, typically in cm2.. The accuracy of the BA is questionable, especially in whole body scans where bold assumptions need to be made, particularly in the upper torso. Areal bone mineral density (aBMD). aBMD is the mineral mass of bone per unit image area in g/em2 Here, a distinction is made between areal density and true volume density. Volume density, the mineral mass per unit bone volume, cannot be directly measured by DXA but can be measured by QCT. aBMD is defined as: aBMD = BMC/BA (g/cm2 ) ‘The following measures are only from whole body scans. Fat mass (FM). DXA FM is the common term used in the DXA field for lipid mass, and is the sum of all lipid mass. Strictly speaking, fat is chemically defined as triglycerides. However, in this book, DXA FM will be defined as all lipid mass, including phospholipids, organ, marrow and subcutaneous adipose. FM is measured in either g or kg. Lean soft tissue mass (LSTM). Bone free, fat free STM is the sum of all soft tissue lean, essentially water, protein, soft tissue mineral and glycogen. It is measured in units of g or kg. Fat free mass (FFM). FFM is the sum of all the non-body lipid, such that: FFM = LSTM + BMC (g) Soft tissue mass: STM is the sum of lean soft tissue and FMs: STM = FM + LSTM (g)Total body mass (TBM). TBM is the equivalent measure to scale weight, typically represented in g or kg. In terms of the above: TBM = FM + FFM = FM + BMC + LSTM. TBM accuracy can be assessed against a calibrated scale, Studies that have investigated the agreement between scale mass and DXA total mass. have found excellent agreement. Per cent fat mass (PCTFM). PCTFM is a region's FM divided by its total mass times 100: PCTFM = FM/TBM x 100 DEXA MEASUREMENT SITES When evaluating bone density using DXA to diagnose osteoporosis, there are several common measurement sites, including the lumbar spine, the proximal hip and the forearm. The standard protocol is to scan two sites, typically the spine and hip. If one of these sites is not available, then the forearm is used. ‘The whole body scan is primarily used for bone mass measurements in children and for body composition measurements in adults. Typical patient doses from DEXA Patient effective doses in DA depend on the type of unit (pencil beam, fan beam, cone beam), the protocol or mode used for the scan (scan area, tube current, scan speed) and the body region being scanned. DXA scans of the forearm are very low, typically less than 1 USv irrespective of the type of scanner and protocol or mode. Lumbar spine, hip or whole body scans each result in an effective dose of about 1 uSv for a pencil beam unit and up to about 10 pSv for a fan beam unit. Some earlier fan beam DXA units initially delivered effective doses as high as 80 uSv. There is not much data for doses from cone beam units, but doses appear to be similar to those for fan beam units. ~ | 5 L| “| 41 PIG 1 active des sriespacde axpsisDXA systems have much in common with other medical X ray imaging systems, with many of the same components. Figure 6 shows a typical X ray gantry for a DXA system including the X ray tube, filtration, pre-patient aperture, examination table or surface, pre-detector aperture and detector. Unlike plain film imaging, but similar to CT systems, the components have a fixed geometry on a gantry even when scanning. The patient lies still while the gantry scans an ROI. Imaging time is typically 3-5 min for FIG. 5. DXA calibration function defined for finding breast PCTFT for pixels of varying thickness on a Hologic Delphi. This curve represented the R values as a composite of fat and fibroglandular breast lean tissue [65]. The black dots are phantom measurements at different thickness and composition. The red lines are the calibration function that was a best fit to the phantom data, For demonstration, horizontal lines show iso-composition and vertical lines show iso-volume (courtesy of J. Shepherd, UCSF). 28 hip and spine scans, and 20 min for whole body. XRAY DETECTOR, DETECTOR COLIMATOR ‘SOFT TISSUE INTENSITY PROFLE FIG 6. Schematic diogram showing the components ofa DXA system (courtesy of. Shephend, csr),Pencil beam systems having only a single detector element have to ‘raster’ scan over an ROI, collecting One pixel at a time. Imaging time is typically 3-5 min for FIG. S. The black dots are phantom measurements at different thickness and composition. The red lines are the calibration function that was a best fit to the phantom data. For demonstration, horizontal lines show iso-composition and vertical lines show iso-volume (courtesy of J. Shepherd, UCSF). 28 hip and spine scans, and 20 min for whole body. Fan beam systems use a linear detector array and collect ten or more pixels at a time. Thus, fan beam systems are much faster than pencil beam systems for equivalent imaging properties. Images of the spine or hip are typically acquired in less than 30 s and 3 min for whole body. Cone beam geometry uses an area detector to take ‘snap-shot’ style images. Although cone beam imaging is the fastest method to take a single energy image, readout time between the images has limited their application in bone densitometry. Ifa subject takes a breath between low and high energy images, severe artefacts result and void the scan. For pencil and fan beam systems, the low and high images of the pixel (pencil beam) or row (fan beam) are collected in milliseconds before the gantry shifts to the next row. Thus, breathing is allowed during the exposure with minimal misregistration. Unlike pencil beam systems, fan and cone beam images do contain X ray scatter; however, scatter rejection is very high for fan beam compared to cone beam systems. t i ' Pencil Beam FIG 7. Pencil and fan beam geometries project the same ROI differently. The pencil beam Image ts projected perpendicular to the plane of the table, whereas the fan beam projection depends on the position of the object within the beam. The projected images. therefor. encompass different physical volumes of tissue when projected back ta the X ray source (courtesy of J. Shepherd, UCSP). GENERATING DUAL ENERGY IMAGES: VOLTAGE SWITCHING VERSUS K-EDGE FILTERING In all DXA systems on the market, the X ray tubes used are standard tungsten anode tubes with focal spot sizes on the order of 0.5 to 1 mm2. However, there are differences in how the dual energy images are created, The two methods in use are voltage switching systems and K edge filtering systems. In a voltage switching system, two X ray tube voltage settings are used to create low and high energy images. The X ray tube power supply switches between a low (70 kVp) and high (140 kVp) voltage setting during alternate half cycles of the power supply. The resulting pulses are very short, 8.33 ms for60 Hz and 10 ms for 50 Hz systems. Copper or brass is used to pre-harden the high energy bear, removing the low energy part of the spectrum and minimizing the overlap between low and high X ray spectra. The filter, voltage switching and detectors are all electronically and mechanically synchronized to sequentially collect low and high energy information for each position of the X ray gantry. Ina K edge filter system, the X ray tube is operated in a steady direct current mode and a K absorption edge filter splits a single X ray spectrum into low and high energy components. The X ray tube kVp is set such that the K edge places a notch in the X ray spectrum that simulates the dual peaked energy spectrum. With ‘an X ray voltage of 100 kV, one of several rare earth filters are used between the patient and the Xray tube by different manufacturers including cerium (Z = 58) and samarium (2 = 62). In these systems, since both high and low energy X rays are intermixed, the energy separation is done at the detector using pulse height measurements. DXA SCAN ACQUISITION DXA REGIONS OF INTEREST ‘There are several ROIs that can be defined, with each having unique information to offer. The optimal site depends on the intent of the scan. For bone density, regions with higher contents of cancellous (high turnover) bone are more sensitive to osteoporotic and treatment changes. However, longitudinal studies suggest that most ROIs currently defined (spine, femur, radius and calcaneus) are useful for predicting general fracture risk (i.e. fractures of any bone). Of the measures available from DXA, BMD, BMC and AREA, the best assessment of risk is BMD [77]. However, the definition of osteoporosis from the World Health Organization (WHO) considers only the BMD of the femur neck. WHO criteria for diagnosing osteoporosis are given in Section 7.2.1. Only the whole body scan mode can measure fat, ean and bone mass. The most common regions and scan modes are outlined below Spine This ROI is the most common for diagnosis of osteoporosis. The scan typically starts at LS and ‘ends at T12. The patient lies supine on the scanning table, with their knees flexed and shins elevated on 2 positioning cube to eliminate lumbar lordosis and flatten the spine against the table top. The aBMD measure of interest is typically for the total of L1-L4 in the postericanterior projection (X ray tube is behind the patient and the detectors over their abdomen). The BMD from this projection includes not only the vertebral bodies but the processes as well. Artefacts and error sources are common for older patients, resulting from extraneous calcifications in the walls of the aorta, deformations from degenerative disc and apophyseal joint disease with its consequent hyperostosis. These types of artefacts typically cause BMD to be falsely elevated [78]. Other conditions that can cause falsely elevated aBMD are vertebral wedge (crush fracture), Paget's disease of bone, sclerotic metastases and haemopoeitic tissue in the marrow of vertebrae. An example PA spine scan is shown in Fig. 8. 32| Beare vor 2 esc8se FIG 8, Example of a DXA Pl spine report, LI-L has been analysed and the total reported. It should be noted thatthe spine ts centred inthe Scan, there is na curvature tothe spine. the iliac crest is slightly visible and there is no twisting, all signs af good Scan acquisition technique (courtesy of J. Shepherd, UCSF), ‘The development of the lateral spine projection was aimed at isolating the vertebral body from the posterior processes to increase the percentage of trabecular bone being analysed. By pairing the lateral scan with the vertebral width from the PA scan, the true volumetric density of the vertebral body can be estimated. The age related change in adults of lateral BMD is higher than with PA spine, and the correlation of volumetric BMD measured by lateral DXA and QCT is high, and stronger than that between PA-DXA and QCT [79]. However, overlap of the iliac crest and L4 and the ribs with Li reduce the typical usable ROI to L2-L3. Furthermore, the precision of lateral BMD is typically worse than PA DXA, such that its ability to monitor change is similar to PA spine. Proximal femur The proximal femur is a common scan site because of the high mortality associated with fractures at this site. In the United States of America, 24% of hip Example of a DXA PA spine report, L1-L4 has been analysed and the total reported. It should be noted that the spine is centred in the scan, there is no curvature to the spine, the iliac crest is slightly visible and there is no twisting, all signs of good scan acquisition technique (courtesy of J. Shepherd, UCSF). 33, fracture patients aged 50 and older die in the year following their fracture. In scanning the proximal femur, the leg is slightly abducted and internally rotated using a positioning device in order to maximize the projection of the femoral neck. Each DXA manufacturer has a different and unique positioner to accomplish this. By not rotating the femur adequately, the femoral neck is foreshortened and this falsely increases the BMD. Positioning of the femoral neck is, therefore, critical to maintaining good precision and comparability to reference data, The ROIs quantified are typically the total, femoral neck, trochanter, intertrochanter and Ward's regions. A typical femur report is shown in Fig. 9.FIG 9. Typical DXA scan report for the proximal femur It should be nouced thar the axis of the shaft és vertical in the image. This image has « fairly pronounced lesser trochanter that oral neck has the appearance of being elongated and, thus, ths patent most likely has a very pronounced less trochanter (courtesy of J. Shepherd, UCSF). could signal that the femur was not fully rotated. However, the fe The specific definition for each ROI is different for each manufacturer. The total femur and femoral neck regions are commonly used for diagnosis, and the trochanter and Ward's triangle regions are seldom used except for research, The regions are shown in detail in Fig. 10. FIG. 9. ret = fret IG 10. Fomor ROD, The tt fomur HOY the um of the te sd regen eer eck, roche tnterrochter) The Maas angle ROY mal rang) ona th te gon on sol ami a hl oma ce of Shp. CSPtypical DXA scan report for the proximal femur. it should be noticed that the axis of the shaft is vertical inthe image. This image has @ fairly pronounced lesser trochanter that could signal that the femur was igated and, thus, this not fully rotated. However, the femoral neck has the appearance of being elon patient most likely has a very pronounced less trochanter (courtesy of J. Shepherd, ucsF). the patient sitting on a chair next to the the hand ina fist and, on some scanners, it may be necessary to scan the port is shown in Fig. 11. Forearm DXA scanning of the forearm is performed with scanner table with the forearm resting on the table top, secured on a positioning board with a restraining strap. In children, forearm at their side while lying on the scanner. A typical DXA scan re ‘onns0080n aL Foren a + aoqnenne vie 0-03 wer |‘ Femwumtora 26200 | ot s (eats [oveconaas 7 saxsizensos5t FIG 11. Typical forearm DXA scan of the right forearm lt shouldbe mote that the forearm is ented in the mage, and thatthe radius and ulna ane straight fthere ts substantial deviation rom the above. the scan shouldbe recur (courtesy of J. Shepherd. UCSE). Forearm BMD measurements are typically reported forthe ultradistal, distal (mi-radius) and shaft (one-third radius) regions. The ultraistal site is useful because it contains the highest percentage of trabecular bone in the forearm. The one-third radius region is useful as a site containing entirely cortical bone. 5.1.4, Total body Total body DXA for bone mineral is of interest because It offers a comprehensive view of total body mineral. This can be useful for calcium balance studies and paediatric studies interested in developmental bone mass. A FIG. 10. Femur ROIs. The total femur ROl is the sum of the three shaded regions (femoral neck, trochanter, intertrochanter). The Ward's triangle RO! (small rectangle) overlaps the other regions and is nota unique area in the total femur (courtesy of shepherd, UCSF). 35 typical whole body DXA scan is shown in Fig. 12.OwtErery i raigger ra pical whole body DXA scan. I should be moted that the fxs have bee gently eld rogether with a Velero strap. the hands are flat and the body centred. straight and completly nahin the scan ficld. If part of the body 1 out of the scan field, some of the techniques herbed in the obesity analysis section must be used (courtesy of. Shepherd, UCSF). FUG. 12 For precise results, itis imperative that the patient be placed on the scanning table in 2 supine position, with all parts of the body, including the arms, included inthe scan field, Total body scans measure BMC dnd average BMD of the total skeleton together. Subregion values are also reported for the skull, arms, tibs, thoracic and lumbar spine, pelvis and legs [82]. In addition to BMC and BMD, total body OXA {quantifies the composition of soft tissue in terms of fat and lean mass [83]. However, OXA cannot solve bone mineral, fat and soft tissue lean mass simultaneously. Thus, in areas where the X ray beam does not intersect bone, itis possible to estimate the masses of fat and lean tissue separately (4]. When bone is present in the pixel with soft tissue, however, only BMD and total (fat and lean) STM can be measured. Extrapolation of measurements of percentage of body fat in soft tissue over adjacent bone means that a whole body DXA scan can provide FIG. 11. Typical forearm DXA scan of the right forearm. It ‘Should be noted that the forearm is centred in the image, and that the radius and ulna are straight. IF there is substantial deviation from the above, the scan should be reacquired (courtesy of J. Shepherd, UcsF). 36 estimates of total body fat and lean mass as well as BMC [5]. The placement of the RO! cut lines are manufacturer specific and the reader should refer to the owner’s manual of interest for placement guidelines. Vertebral fracture assessment Due to the relatively high resolution of fan beam DXA scanners, ‘anatomical details of the examined region are depicted clearly. Using DXA to obtain lateral images of the lumbar spine allows the scanning beam — in contrast to conventional cone beam radiography — to be generally parallel to the vertebral endplates (Fig. 13). These images can be acquired as either dual orsingle energy for better visualization. The single energy acquisition reduces X ray noise. The FIG. 12 Typical whole body DXA scan. It should be noted that the toes have been gently held together with a Velcro strap, the hands are flat and the body centred, straight and completely within the scan field. If part of the body is out ofthe scan fed, some ofthe techniques described inthe obesity analysis section must be used (courtesy of J. Shepherd, UCSF). 37 FIG. 13, LUG 13 Lateral vertebral assexyment ts used to better visualize vertebral fractures, The left ‘rapresentation The dal energy view of the same spine is shown om ‘classified wing scaring methods reflecting the severty of the marge ts the single the right, Fractures can be fracture. Lateral vertebral assessment is used to better visualize vertebral fractures. The left image is the single energy representation. The dual energy view of the same spine is shown on the right. Fractures can be classified using scoring methods reflecting the severity ofthe fracture. 38 dual energy image reduces soft tissue artefacts. Several terms used in reference to the DXA approach, including vertebral fractureassessment, morphometric X ray absorptiometry [84], lateral vertebral analysis and instant vertebral analysis, essentially mean the same thing. In general, vertebral fracture assessment allows for better definition of vertebral dimensions than conventional X rays. Before scanning the patient Before patients come in for their scans, they should be informed of the following issues: (2) It should be ensured that patients can tolerate lying flat on their back and keeping still for at least 10 min. 2. The weight of subjects should be determined. All DXA systems can scan patients up to 300 Ib (136 kg) If they are over 300 Ib, they may need to have an alternative bone density or body composition test performed. The DXA system owner's manual should be checked for the specific limits of the system. (3) It should be determined whether the subjects have had any medical imaging procedure recently. If they have received contrast, such as barium or gadolinium, they should be scheduled two weeks after contrast was administered. (4) If patients are premenopausal, they should be asked whether there is any possibility that they might be pregnant. In some clinics, a pregnancy test may need to be administered before the examination. Patients should be informed of this possibility. (5) Calcium tablets should not be taken in the 24 h before the examination. inimizes the (6) Patients should wear comfortable, loose fitting clothes, such as a sweat suit — this need to change into a hospital gown. (7) Patients should avoid wearing clothing with metal components such as zippers, underwired bras or rivets. (8) Patients should be asked whether they have had a prior bone densitometry test. If so, the patient should bring those test results with them. (9) Patients should be asked to bring in the appointment information and their referring doctor's contact information if applicable. (10) For body composition studies, patients should be scanned in the morning after a 12h overnight fast for consistency. ‘On the day of the examination, before subjects are scanned, the following should be checked: (1) Patients have complied with the recommendations listed above. Specifically: (2) Patients should be asked again whether they have had any medical procedures in the preceding two weeks, such as CT or MRI. If they have received any contrast (barium, gadolinium, etc.) they should wait at least two weeks before their DXA scan. (6) The menopausal status should be re-checked and whether a pregnancy test or question relating to possible pregnancy has been administered.(2) Subjects should be dressed in a hospital gown or scrubs, wearing only underpants and, if necessary, thin socks. A thin sheet may be placed over subjects for warmth. (3) All radio-opaque objects should be removed from the scan area (underwired bras, jewellery, belts, etc.). (4) When performing follow-up scans, the previous image of the baseline scan should be printed out to ensure duplicate positioning and scan parameters. Scanning the patient Positioning is, by far, the most common limiting factor to precision. Phantom scans, with no repositioning can commonly have an aBMD imprecision of 0.5%. For PA spine scans, the imprecision in vivo for the same measurement site is typically 1~1.5% because of the errors associated with projecting the patient's bones slightly differently. Patient movement during the scan causes the bone and soft tissue projection to change, slightly altering the projection. Changes in positioning between the baseline and follow-up can be difficult to detect. This error, as well as the imprecise placement of the ROI cut lines, can usually be minimized by training. Note: When scanning subjects, itis important to keep in mind that itis much less time consuming to re- scan the subject immediately if a problem is detected, rather than having to recall the subject for a repeat of the scan on another day. The following is a systematic method for positioning patients: (1) The same scan mode should always be used throughout the patient's baseline and follow-up visits (ie. ‘array’ versus ‘fast array’, ‘thick’ versus ‘standard’, etc.) (2) Keeping the scan width and length set to the default settings is preferred in most situations. (3) For PA lumbar spine scans, it may be necessary to use a positioning block to remove the lordosis: from the lower back. The spine must be straight and centred in the scan field. (4) For femur scans, the femur must be rotated and held in position with a positioning device. There are also dual hip scanning protocols that position both hips simultaneously. However, the positioning device may need to be adjusted between scans to scan both hips with correct abduction. (5) For forearm scans, the patient is scanned sitting in a chair without wheels. This may be difficult for people of short stature. For children, it may be more appropriate to have them lie prone on the table with their forearms extended above their heads or supine with the arm of interest away from their side. jer end towards the foot end of the table in order This may require the patient to be positioned with ei to acquire the correct projection using the correct (left/right) scan mode. (6) Whole body is the most difficult with the most technical challenges. For whole body positioning, please refer to Fig. 14. The subject should be positioned in the centre of the table, aligned with the long axis of the scanner, with their head near the head end of the table. The subject's head should face straight up, not turned to the left or right. If required for subject comfort, only radiolucent pillows should be used. If pillows are used, however, a note should be made to use the same pillow again during follow-up measurements: (a) The legs and feet should be positioned together with a Velcro strap around the ankles to help avoid movement. Feet should be kept relaxed with the toes pointed upwards (Fig. 15). (b) Hands should be positioned with palms flat against the scan table. Space should be maintainedbetween the arms and the torso when possible. Ifnecessary, may be placed in a lateral position next to the hips. Hands sh them in the scan field. if necessary, the subject's hands should be taped to the scan table. For patients who are too tall to fit within the scanning limits, itis acceptable for the feet to extend beyond the lower scan limit line. The knees should not be bent to keep the feet within the scan field, (7) For all scan modes, the scan should be monitored durin or the subject moves, scan restarted. with larger or heavier subjects, the hands }ould not be tucked under the hips to keep 1g acquisition. If the positioning is not correct etc,, the scan should be aborted, the subject repositioned if necessary and the (8) The patient should not be engaged in conversation because it may cause motion, but encouragement. and updates on scan progress should be given. (2) After the examination, before the patient has left the table, it should be verified again that no ‘movement has occurred during scanning. If the scan is not correct, the patient should be re-scanned