Introduction

Diagnostic characterization of the equine musculoskeletal system typically relies on a combination of imaging modalities. Conventional radiography for assessment of bone and ultrasonography for assessment of soft tissues are currently the standards of initial diagnostic assessment of musculoskeletal injuries in horses.¹ However, the sensitivity of those diagnostic imaging techniques for detection of subtle pathological lesions is limited, compared with volumetric imaging modalities such as CT and MRI; the latter 2 techniques can be used for more comprehensive identification and characterization of pathological lesions. Disadvantages of volumetric imaging include cost, the need for patient anesthesia in some instances, prolonged scan times, and provision of dedicated space for equipment installation. The ideal method for diagnostic imaging of the equine musculoskeletal system would be inexpensive and portable, produce real-time images that are easy to obtain and yield detailed anatomic information regarding various tissue types (thereby enabling accurate assessment of pathological lesions), and provide findings complementary to those obtained with radiography and ultrasonography.

Tomography-based techniques were initially developed in the 1950s to reduce the effects of anatomic overlap on interpretation of conventional radiographic images.² Despite the high radiation dose, the technique persisted even after the advent of CT for applications such as IV pyelography. More recently, with the advent of advanced reconstruction techniques, digital tomosynthesis has been commonly used for mammography^3,4 and musculoskeletal imaging of knees,^5,6 wrists,⁷ and hands^8,9 in humans. Digital tomosynthesis uses radiographic images obtained at different angles within a single plane to create multiple slices of a structure of interest, with a radiation dose that is markedly less than that associated with CT. This image set helps to overcome the problem of masking caused by superimposition of structures—a frequently encountered limitation of radiography.^10,11 The potential ability to rapidly obtain 3-D information by means of modified radiographic technology makes tomosynthesis attractive as a potentially useful addition to routine clinical diagnostic imaging procedures. The objective of the study reported here was to investigate the technique of tomosynthesis for imaging of the distal portion of equine forelimbs. The intent was to evaluate the strengths and weaknesses of tomosynthesis with regard to characterization of the different tissues of the metacarpophalangeal joint (MCPJ). The appearance of pathological lesions of MCPJs in images obtained by tomosynthesis and radiography were compared, with assignment of MRI as the imaging standard for articular cartilage and periarticular soft tissues and CT as the imaging standard for bone and bony margins. We hypothesized that tomosynthesis would provide a more complete characterization of the MCPJ, compared with that achieved radiographically for lesions in a specific plane, but would not provide as a comprehensive an assessment as CT or MRI, and that tomosynthesis could be validated as a volumetric imaging technique with potential benefits for diagnostic assessment of equine MCPJs.

Materials and Methods

The study used a cross-sectional method-comparison design. Cadavers of horses euthanized by an IV overdose of pentobarbital sodium for reasons unrelated to the study were obtained, and the distal portion of 1 forelimb was collected from each cadaver. The MCPJ of each limb underwent separate imaging studies using radiography, CT, MRI, and tomosynthesis. After imaging, MCPJs were disarticulated and evaluated for the presence of gross lesions in the bones, articular cartilage, and periarticular soft tissues of the joint. Gross evaluation was used to confirm CT and MRI observations and as a reference category for imaging comparisons. The size and shape of the lesions were also noted for comparison with imaging findings.

Tomosynthesis

A compact benchtop tomosynthesis device, constructed from existing equipment for this preliminary feasibility study, was used to obtain images of the distal portions of the 4 equine limbs. The system was comprised of a carbon nanotube x-ray source array and an off-the-shelf flat panel detector (C7940DK-02; Hamamatsu Photonics KK; Figure 1). The linear source array contained 7 individually addressable x-ray focal spots, each approx 1 × 1 mm, that spanned approximately 85 mm. Originally designed for human intraoral tomosynthesis imaging,¹² the source array had a maximum anode voltage of 70 kVp and a maximum tube current of 10 mA. The system was constructed for this feasibility study with a source-to-detector distance of 40 cm, resulting in an angular span of 12°.

View Full Size

Photographs of a custom-built compact tomosynthesis device and flat panel detector used to obtain images of the MCPJ of 4 forelimbs from equine cadavers. A—In this image, the source array and detector set up for cadaver limb scanning with schematic representation of x-ray beams are shown. B—The distal portion of a cadaver limb is positioned for a dorsal scan of the MCPJ. The cadaver limb is supported in a custom-made holder, which slightly increases the object-to-detector distance, but allows scanning of limbs at a fixed position.

Citation: American Journal of Veterinary Research 82, 11; 10.2460/ajvr.82.11.872

• Download Figure
• Download figure as PowerPoint slide

Each limb was positioned for the tomosynthesis scan by an experienced board-certified veterinary surgeon (CEK) in a mounting fixature to ensure clinical relevance. The limb was oriented with the long axis either parallel or perpendicular to the source array. A tomosynthesis scan consisted of electronic triggering to the source array to produce 7 individual basis projection images at 7 discrete angulations with respect to the limb. Images were acquired with no mechanical motion. These projection images were processed in a reconstruction algorithm to generate a quasi-3-D image stack parallel to the detector plane. A total of 80 to 100 reconstruction slice images were generated for each scan in a given plane, each representing 1 mm of object thickness. Images for each MCPJ were acquired in multiple views, including lateromedial, dorsopalmar, dorsolateral-palmaromedial oblique, and dorsomedial-palmarolateral oblique views. Images were obtained with the following settings: anode voltage, 70 kVp; anode current, 7 mA; exposure per source, 80 milliseconds (total exposure for 7 sources, 560 milliseconds); exposure for the tomosynthesis scan, 3.92 mAs; source-detector distance, 400 mm; angular span, 12°; and estimated entrance dosage, 1.7 mGy. Total scan time per limb orientation (eg, dorsopalmar) was 15 seconds.

Results

Identification of MCPJ lesions by gross evaluation and by radiography, tomosynthesis, CT, and MRI

Among the 4 equine cadaver forelimbs used in the study, limb 1 appeared normal with no grossly visible lesions with an intact surface of the articular cartilage on the distal aspect of MC3. Limb 2 also appeared grossly normal. Limb 3 had marked palmar osteochondral disease characterized by subchondral bone erosion and extensive articular cartilage loss. Limb 4 had an incomplete, nondisplaced chronic fracture of the medial PSB. Both CT and MR evaluation of limb 2 revealed a focal subchondral fissure associated with a fissure lesion in the medial condyle of MC3. For each of the 4 limbs, accurate lesion detection in images obtained with tomosynthesis, CT, or MRI was achieved. Radiographic lesion detection was correct for 3 of the 4 MCPJs; the focal subchondral lesion in limb 2 was not detected radiographically (Figure 2).

View Full Size

Comparative dorsal plane images of 1 of the 4 cadaver limbs (limb 2) obtained with radiography (A), CT (B), and tomosynthesis (C). A subchondral fissure bordered by sclerosis (arrows) is detectable in the medial condyle of the distal aspect of MC3. This lucent area was observed on images obtained with tomosynthesis and CT but not evident on images obtained with radiography.

Citation: American Journal of Veterinary Research 82, 11; 10.2460/ajvr.82.11.872

• Download Figure
• Download figure as PowerPoint slide

In images obtained by radiography and tomosynthesis, the appearance and size of observed lesions for limbs 1, 3, and 4 were similar. With the exception of limb 2, the extent and physical characteristics of the lesions observed in images obtained by the 4 techniques were similar to what was observed on gross evaluation of the joints. For limb 2, the focal subchondral lesion was most prominent on CT and MRI images and was minimally visible on evaluation of the articular surface of the joint.

Discussion

The study of the present report involved a comparative assessment of radiography and a proof-of-concept benchtop tomosynthesis system for evaluation of a limited number of equine MCPJs. Results indicated that tomosynthesis can provide valuable information regarding bone status and allowed better detection of subtle lesions of the subchondral bone, compared with radiography. Evidence in support of this conclusion was provided by the ability of tomosynthesis to detect a single, radiographically occult lesion in one of the MCPJs. Although the level of detail for bone and articular cartilage provided by tomosynthesis was not equivalent to that achieved by CT and MRI, respectively, application of tomosynthesis as an adjunctive, rapid imaging modality for examination of the distal portions of equine limbs may be valuable.

Radiography and tomosynthesis both use x-ray projections to obtain images of a given region of interest. Intuitively, multiple projections acquired within a single plane by tomosynthesis have the potential to provide more information than a single image acquired by radiography. Tomosynthesis improves spatial resolution in the x-y plane, thereby reducing the effect of superimposition of structures, which is commonly observed with radiography. Reduction of that effect may enable identification of more subtle lesions, as found in the present study. The subchondral lesion with surrounding sclerosis in the MCPJ of limb 2 was readily detectable with tomosynthesis, but owing to its location on the distal articular margin of MC3 and superimposition of sclerotic bone tissue, it was not visible with conventional radiography. A small fissure in the articular cartilage was visible on gross evaluation of the limb, but that gross finding underrepresented the degree of change within the subchondral bone, as determined by CT. The radiographic images were reviewed after scoring was completed, and this pathological change was not visible radiographically even with prior knowledge of the location and characteristics. For each tomosynthesis scan, 80 to 100 reconstruction slice images (each representing 1 mm of object thickness) in a given plane were generated. In reality, the tomosynthesis reconstruction algorithm can generate a variable number of slice images to represent an object volume.¹⁴ Given the small number of limbs evaluated in the present study, the difference in lesion detection between radiography (3/4 limbs) and tomosynthesis (4/4 limbs) should not be overstated. Despite this comparatively improved lesion identification with tomosynthesis, the differences between radiography and tomosynthesis with regard to evaluation of subchondral bone and adjacent trabecular bone were not significant. As a technique, tomosynthesis is an extrapolation of radiography, so we would expect the appearances of lesions with the 2 modalities to be similar; however, given the improved ability of tomosynthesis to reduce superimposition, we would have expected to find significant differences in detection of lesions and the appearance of subchondral bone between these modalities. A notable limitation in the design of the present study was the examination of limbs with lesions that were distinct enough to be seen on radiographs; further studies should be focused on evaluation of tomosynthesis for the detection of subtle lesions, especially those associated with subchondral bone. Tomosynthesis and conventional radiography for evaluation of features of osteoarthritis in human knees have been compared.⁵ Osteophytes and subchondral cysts were more accurately identified with tomosynthesis than with conventional radiography; furthermore, persons with lesions identified by tomosynthesis were more likely to report pain than were persons without lesions identified by tomosynthesis.⁵ Osteophytes and subchondral bone cysts were not specifically present in the equine limbs evaluated in the present study, but further research is needed to determine whether these findings are applicable across species. Similar to humans, the health of subchondral bone has a direct impact on the health of joints in horses,^15,16,17 and development of noninvasive imaging methods for evaluation of this specialized bone would allow clinicians to have a more complete understanding of many commonly observed injuries. Injury to subchondral bone in athletic horses may result in signs of pain and lameness, predisposing affected joints to fracture or inciting or perpetuating osteoarthritis.^18,19,2021 Detection of such lesions with tomosynthesis may enable early intervention and reduce the likelihood of catastrophic injury in horses.

Results of the present study indicated that high-quality diagnostic images of equine MCPJs can be obtained with tomosynthesis. The tomosynthesis system used in this study was for research purposes only, but was substantially smaller than the commercial tomosynthesis systems used for mammography. The tomosynthesis system used 7 projection images and a 12° angular span, comparatively fewer images and less angular span than what are used in orthopedic tomosynthesis studies in humans.²² Increasing the number of projection images or angular span may have provided additional information for specific regions of interest in equine MCPJs, but an objective comparison of different system configurations for equine imaging evaluation was outside the scope of the present study. The source array used in the present study was also originally fabricated for a dental application, with a maximum anode voltage of 70 kVp. Any future purpose-built system would need to incorporate a maximum anode voltage of 80 to 90 kVp to be appropriate for tomosynthesis of the distal portion of equine limbs. Such a purpose-built system would likely contain a source array with additional focal spots with smaller focal spot size for increased system resolution and provide increased angular coverage, as well as use a larger, flat panel detector. Thorough optimization studies would be required to balance image quality with the physical size of the tomosynthesis system.

The tomosynthesis system used in the present study was configured so that the source array required no mechanical motion or complex gantry setup to acquire the projection images, enabling the hardware to be compact and relatively simple, compared with other commercially available tomosynthesis systems. This configuration allowed the system to be packaged into a portable footprint, with the potential to be configured for imaging of standing horses in a clinic or field setting. Although not assessed in this study, there is a commercially available portable tomosynthesis system (Equetom; Universal Medical Systems Inc) for imaging of horses; on the basis of experiences of one of the authors (KTS), initial clinical impressions regarding use of the system for obtaining images of horses’ heads and distal limb regions are promising. The focus of the present study was to evaluate the capabilities of tomosynthesis technology with regard to equine limb assessment, which could have been strengthened by evaluation of the commercially available portable tomosynthesis unit in live horses. The commercially available portable tomosynthesis unit also has the capability to perform digital radiography of equine limbs and has a similar system setup to those of portable digital radiography systems, wherein the evaluated horse is sedated, standing, and appropriately positioned to stand squarely, and the plate and generator are placed around the area of interest on the distal portion of the limb. The availability of a single portable system that can be used for radiography and tomosynthesis in the field emphasizes the relevance of considering these techniques together. Although planar imaging techniques, such as radiography, are inferior to volumetric imaging techniques, such as CT and MRI, they remain the mainstay preliminary assessment methods for many equine veterinarians.

Another clear advantage of tomosynthesis is the rapid image acquisition. Similar to radiography, with which images can be acquired in < 1 second, tomosynthesis can acquire many images rapidly. In the present study investigating the distal portion of equine forelimbs, all images within a single plane were acquired in an approximately 15-second period, whereas the duration of CT and MRI examinations was approximately 5 and 30 minutes, respectively. The evaluated tomosynthesis system with a 12° angular span was an early prototype, which has subsequently been modified to perform a scan with a 40° angular span in approximately the same time (a scan with a 12° angular span requires approx 1 second to perform²³). This adapted system would be expected to improve lesion conspicuity. Furthermore, despite advances in artifact reduction software, metal implants in the distal portion of an equine limb create numerous artifacts on both CT and MRI images, making image interpretation more challenging. Tomosynthesis appears to be less susceptible to metal artifacts in images, compared with CT, and postprocessing techniques have been reported²⁴ to mitigate these artifacts and improve image quality, expanding the potential clinical applications of tomosynthesis. The use of synthetic radiographs, which are generated from tomosynthesis reconstruction data, is a rapidly emerging technique that reduces the total radiation dose to patients that would have otherwise undergone tomosynthesis and radiography.²⁵

When considering a clinical equine case, tomosynthesis could be used as a complementary imaging modality with radiography. For example, once the region of lameness has been localized with diagnostic analgesia, a standard series of radiographic views could be obtained. If a lesion is identified on radiographic images—even within a single plane—then additional tomosynthesis images could be obtained to further characterize a detected abnormality. Additional diagnostic imaging recommendations could be made on the basis of the combined findings of radiography and tomosynthesis. Alternatively, if clinical lameness has been localized in a horse but no lesions are observed with radiography, tomosynthesis may be used in multiple planes in an attempt to identify and characterize a suspected lesion. Future tomosynthesis studies could be directed toward assessment of the navicular apparatus of horses. In the present study, images of the middle and distal phalanges of 2 equine cadavers were obtained but not specifically scored or included in the evaluations. Previously, Johnson et al²⁶ demonstrated that conspicuity of lesions along the palmar or plantar cortex of the navicular bone is directly impacted by the projection angle of radiographs. It is our impression that tomosynthesis will aid in identification and characterization of these types of lesions, and further clinical investigation is needed to confirm these assumptions.

The study reported here was an initial assessment of tomosynthesis technology for acquiring images of the distal portion of equine limbs. There is an inherent challenge in attempting to obtain a limited number of images within a single plane for creation of a pseudo-3-D image that mitigates planar superimposition. It is not surprising then that there are notable differences between tomosynthesis and CT with regard to evaluation of bone that cannot be overcome by collection of multiple images across a given plane. Results of the present study also confirmed that articular cartilage cannot be appropriately assessed with a tomosynthesis system that has a 12° angular span and that, of the 4 assessed imaging techniques, MRI was the superior imaging method for cartilage evaluation. However, MRI is not the only imaging modality for evaluation of articular cartilage because CT performed with appropriate contrast resolution does allow some evaluation of the articular cartilage. Although not specifically addressed in this study, there is a learning curve for evaluation of images obtained with tomosynthesis; instead of a single image, as obtained with radiography, the evaluator is required to scroll through multiple images within a given plane. Although assessment of images obtained with tomosynthesis is less affected by superimposition of structures, compared with assessment of radiographic images, confidence in the use of this modality requires experience. If a lesion is suspected, it is imperative to center that potentially affected region of interest in the tomosynthesis beam to reduce the effect of data truncation and improve the evaluator’s ability to characterize the lesion. Tomosynthesis generates reconstruction volumes in the shape of a truncated pyramid. As the angular span increases, it becomes more critical for the suspected lesion to be centered in the field of view because the effective angular coverage is decreased for thick objects near the detector’s edge. The impact of data truncation on digital breast tomosynthesis has been discussed,^27,28 and when considering the use of this technology on horses, it is important that the operator has a working knowledge of the more likely locations of lesions (eg, the palmar aspect of the surface of the MCPJ). Another consideration is that a tomosynthesis reconstruction enhances anatomic features of a region that may be present on 1 or more projection images. If the goal is to visualize a joint space or articular surface, 1 or more of the projection images should be oriented to allow the beam to traverse parallel to this region of interest. This is similar to the approach described for radiography, in which the image contrast is improved when the x-ray beam is properly angulated and the limb is positioned to open the joint space or expose the lesion. The importance of proper positioning of the anatomic region of interest and beam angle has been well described in the equine medical literature.^29,30,31,32

Tomosynthesis generally offers some flexibility in patient positioning owing to the acquisition of data from multiple angles. In the present study, the angular span and number of projection images were lower than typical reported values for human tomosynthesis systems,³³ and the detector area (12 X 12 cm) was relatively small, posing additional challenges to positioning. Arguably, the largest limitation of tomosynthesis is that the information generated is most helpful when a specific plane of interest is examined. Although it would be possible to obtain tomosynthesis images in multiple planes, the technique is likely best used to further delineate a suspected pathological lesion observed on radiographs. Thus, the need for radiography is not negated by availability of tomosynthesis, but the information obtained with each modality should be considered complementary.

In the present study, the median scores for subchondral bone determined for radiography and tomosynthesis were compared, and an a posteriori power calculation indicated that a total of 40 limbs (selected randomly to avoid the potential bias of including lesions identified with radiography) would be required in a blinded, multievaluator study to determine whether a true difference exists between radiography and tomography with regard to evaluation of the equine MCPJs (including the percentages of false-positive and false-negative findings, if any) and validate the observations from the present study. The authors chose to use CT and MRI as the standards for comparison between imaging modalities; however, gross evaluation was used as the ultimate reference for the presence of lesions visible on the articular surface. The findings for limb 2 in the present study indicated that, in some instances, the extent and nature of subchondral bone lesions may be more thoroughly represented with volumetric imaging than through observation of the articular surface. Histologic examination of subchondral bone specimens would have been required for a gold-standard assessment of these lesions.

Overall, findings of the present study indicated that tomosynthesis can be used as a diagnostic imaging modality for evaluation of equine MCPJs. In this limited study, tomosynthesis and radiography were comparable with regard to evaluation of subchondral bone and identification of adjacent trabecular bone borders; however, tomosynthesis was slightly inferior to radiography with regard to characterization of periarticular margins. Importantly, for one of the evaluated limbs, tomosynthesis was more accurate than radiography with regard to identification of a pathological lesion in the subchondral bone, which could have been obscured because of the superimposition of bone and soft tissue structures on radiographs. Further investigation is needed to determine whether identification of nonradiographically detectable lesions with tomosynthesis is a repeatable strength of this modality. The ability to obtain sequential images of a structure across a single plane with tomosynthesis suggests that this may be an adjunctive technique that is complementary to other diagnostic imaging modalities. As determined in the present study, tomosynthesis appears to have promising applications for clinical use in horses and warrants further investigation. Orthopedic studies in horses will be necessary to determine whether tomosynthesis will provide image quality and diagnostic accuracy comparable to that reported for orthopedic tomosynthesis in humans.