British Journal of Radiology (2005) 78, 587-595
© 2005 British Institute of Radiology
doi: 10.1259/bjr/29830482
3D reconstruction of the skeletal anatomy of the normal neonatal foot using 3D ultrasound
C J C Cash, MCRP, FRCR
1
G M Treece, MA, PhD
2
L H Berman, MRCP, FRCR
1
A H Gee, MA, PhD
2 and
R W Prager, MA, PhD
2
1 University Department of Radiology, Addenbrooke's Hospital, Hills Road, Cambridge CB2 2QQ and 2 Cambridge University Department of Engineering, Trumpington Street, Cambridge CB2 1PZ, UK
 |
Abstract
|
|---|
Currently imaging plays a limited role in the assessment of the neonate with a foot deformity. The aim of this study was to establish a technique for examining the neonatal foot with three-dimensional ultrasound (3D US). 3D US was attempted on the normal feet of 20 infants (9 male, 11 female) under 6 weeks old (range 35–41 days). The data sets were obtained whilst the infants were feeding or asleep to minimize movement artefact. A high-resolution optically tracked freehand 3D US system (Diasus, 16 MHz transducer) was used with Stradx software to acquire and analyse the data sets. Manual segmentation of the non-ossified tarsi from the data sets was performed. Five infants were too restless to be examined. 107 data sets were recorded from 22 feet of the remaining 15 infants. 21 of the data sets were discarded due to movement artefact. 86 were suitable for manual segmentation. Surface interpolation of the segmented data sets produced surface rendered reconstructions illustrating the complex 3D anatomy of the foot. This new technique may offer a method of examining the deformed foot, e.g. congenital talipes equinovarus.
|
Introduction
|
|---|
Classical anatomical texts provide detailed descriptions of the adult foot based upon cadaveric dissections and cross-sectional imaging. There are few detailed anatomical descriptions or studies of the normal neonatal foot and even fewer imaging studies of the normal neonatal foot, perhaps because imaging currently plays a very minor, or no role in the assessment of the child with a congenital foot abnormality. Orthopaedic surgeons rely on clinical examination of the foot to grade the deformity. Plain radiographs may be used in this assessment [1], however, their use is limited by the lack of ossification of the tarsus.
MRI has been used in research to investigate foot deformities [2–4] but frequently requires sedation as well as being a lengthy method of imaging. Whilst the lack of ossification limits the use of ionizing radiation, the non-ossified tarsus may be completely imaged by modern ultrasound, with increasingly improved resolution. Ultrasound is safe and relatively inexpensive, when compared with CT or MRI. Two-dimensional (2D) ultrasound has been used in the pre-, peri- and post-operative assessment of infant feet with congenital talipes equinovarus (CTE) [5–7].
Using three-dimensional ultrasound (3D US) data sets, sections of skeletal anatomy in multiple planes and surface rendered reconstructions can be produced. A method of examining the normal neonatal foot may prove useful in investigating the abnormal neonatal foot, for example congenital talipes equinovarus and vertical talus. Thus the aim of this study was to establish a method of examining the neonatal foot with 3D US in order to reconstruct the 3D skeletal anatomy.
|
Method
|
|---|
The study design was reviewed and approved by the local regional ethics committee. 20 infants (9 male, 11 female) were recruited into this pilot study. They were all under 6 weeks old (range 35–41 days, mean 40 days) and were attending the ultrasound department for routine ultrasound screening of their hips for the usual clinical indications (e.g. breech delivery, a family history of developmental dysplasia of the hip, or the clinical finding of a positive instability test). Parents had been sent information prior to their attendance and upon arrival were asked whether they would be willing to participate and informed consent was taken. All infants had clinically normal feet and subsequently were found to have sonographically normal hips.
The infants were disturbed as little as possible, and in most instances, the 3D US was performed whilst they were asleep or whilst feeding. Warmed ultrasound gel was used. These efforts were to ensure that the child remained as immobile as possible. The ultrasound examination was performed and the data sets analysed by the same operator (CJCC). The 3D data sets were obtained from free-hand 2D grey scale images from a high-resolution Diasus ultrasound machine (Dynamic Imaging, Livingston, Scotland, UK, http://www.dynamicimaging.co.uk/) using a 16 MHz transducer. The transducer position was continually tracked using optical sensors (Polaris; Northern Digital Inc., Waterloo, ON, Canada, http://www.ndigital.com/). The B-scans were transferred to an 800 MHz PC via an Ethernet link, where they were analysed using Stradx software. This software was written by Cambridge University Engineering Department and is currently contributing to several clinical research projects [8]. It is freely available on the web (Stradx, http://mi.eng.cam.ac.uk/
rwp/stradx) and has been described elsewhere [9]. The system relies on the examined object being completely immobile for accurate position sensing.
For examination of the hind foot, the transducer, at a depth setting of 3 cm, was moved in a linear craniocaudal direction from the medial aspect of the lower calf down over the medial malleolus and across the medial aspect of the hind-foot. The transducer was orientated such that its long axis was parallel to the long axis of the talus (Figure 1a
). For examination of the forefoot the transducer, at a depth setting of 2 cm, was placed over the medial malleolus and moved linearly towards the toes (Figure 1b). The individual B-scans from the data sets (typically 300–400 images per data set) were viewed on the PC. Multiplanar reformats (MPRs) from any projection within the volume of the data set could be chosen and displayed in real time until the desired plane of reformat was achieved (Figure 2). An algorithm within the Stradx software [10] was applied to the data set to correct for any misregistration caused by operator-induced variations in transducer pressure or physiological tremor. The individual cartilaginous anlages of the tarsal bones were identified. From representative B-scans through the data set, the outline of the tarsal structures was manually segmented, producing a series of outlines of the cartilaginous components of the ankle and foot (Figure 3). A surface was interpolated from these outlines [11] to produce a surface rendered reconstruction of the skeletal components of the foot.
View larger version (10K):
[in this window]
[in a new window]
|
Figure 1. The position and movement of the transducer over the medial aspect of the foot for (a) a hindfoot and (b) a forefoot examination.
|
|
View larger version (77K):
[in this window]
[in a new window]
|
Figure 2. The full data set is shown on (a) the left, each B-scan is represented by a white "soccer goal-post" shape. A sagittal plane through the data set is chosen (arrows) and the reformatted data (dashed border) is demonstrated on (b) the right against one of the original B-scans (solid border). a. talar head, b. navicular, c. medial cuneiform, d. intermediate cuneiform, e. lateral cuneiform, f. first metatarsal, g. approximate position of the calcaneum.
|
|
View larger version (48K):
[in this window]
[in a new window]
|
Figure 3. The full data set is shown on (a) the left. One of the manually segmented B-scans at the level of the lower tibial epiphysis/proximal talar body is illustrated in (b). The complete series of segmented outlines is demonstrated in (c).
|
|
The data set took between 5 s and 8 s to acquire, but was repeated up to 10 times for each subject, depending on individual subject movement. Data sets acquired during any detectable foot movement were immediately discarded. Provided the baby remained settled and parental time allowed, both feet were examined. The analysis and segmentation took, on average, 30 min to complete.
|
Results
|
|---|
Subjects
The procedure was abandoned in 5 of the 20 neonates examined because the child became restless before or during acquisition of the first data set. Of the remaining 15 infants, 7 remained asleep long enough to allow both feet to be examined. In the other eight cases, the child was roused by the examination limiting data acquisition to a single foot. A total of 107 data sets were recorded from 22 different feet. Following analysis, 21 data sets were deleted due to minor foot movement that had not been apparent at the time of data acquisition. On average four successful data sets were recorded per foot. 86 data sets were subsequently segmented.
Imaging and analysis
A full 3D data set of the lower tibia, fibula, ankle and hind foot is shown in Figure 2a
demonstrating the movement of the transducer over the ankle and foot. Figure 4 demonstrates the resolution of the individual B-scans. The resolution of a MPR through the whole data set from a forefoot examination is shown in Figure 5. Any operator-induced misregistration was satisfactorily corrected (Figure 6). The small bones of the foot that are completely cartilaginous at birth (navicular and cuneiforms) were easily identified and manual segmentation could be performed satisfactorily. The hyaline cartilage of the non-ossified skeleton appears on ultrasound as poorly reflective structures with an even less reflective margin corresponding to the articular cartilage [12]. This hyporeflective border helps in distinguishing their outline and aids the segmentation process. Occasionally the border of one tarsal bone was difficult to distinguish from an adjacent bone, such as the distal end of the talar head adjacent to the concave proximal border of the navicular (Figure 7a). By reformatting the data in the sagittal and axial plane, and using reference landmarks (Figure 8) it was possible to segment the proximal navicular from the distal talus on the original image (Figure 7b). A similar example is illustrated in Figure 9, where by reformatting the data helps to distinguish the superior surface of the talar dome from the articular margin of the distal tibial epiphysis. Alternatively, improved border delineation could be achieved using narrow band maximum intensity projection [13]. In this technique, the maximum image intensity from a stack of reformatted images covering a short distance (e.g. 1 mm) are displayed (Figure 10a,b).
View larger version (51K):
[in this window]
[in a new window]
|
Figure 4. Illustration of the high resolution of the images. The distance between the two white arrowheads is 4.6 mm representing the sagittal dimension of the navicular.
|
|
View larger version (32K):
[in this window]
[in a new window]
|
Figure 5. (a) An axial reformat through the foot from a forefoot data set. Annotated illustration of the reformat is shown in (b).
|
|
View larger version (75K):
[in this window]
[in a new window]
|
Figure 6. (a) A sagittal multiplanar reformat (MPR) through the foot. (b) Operator induced misregistration can be corrected after acquisition. (c) The anatomy and the plane of the MPR.
|
|
View larger version (129K):
[in this window]
[in a new window]
|
Figure 7. (a) The original B-scan through the talo-navicular joint. Manual segmentation of the talus (dashed line) from the navicular (solid line) is shown in (b), but would have been difficult without the aid of multiple reformats.
|
|
View larger version (77K):
[in this window]
[in a new window]
|
Figure 8. (a) The original B-scan at the talo-navicular joint. Sagittal and axial reformats are shown in (b) and (c), respectively. The white arrowhead in the sagittal reformat (b) indicates the dorsal aspect of the navicular (N). Its corresponding position on the original image (a) is also shown as a white arrowhead. A reformat in the axial z-plane (c) shows the mediolateral extent of the navicular (white arrow). Landmark X indicates the lateral aspect of the navicular shown on the reformatted image (c) and original image (a). These reformats and reference points were used to segment the talus (T) from the navicular (N) on the original B-scan as accurately as possible.
|
|
View larger version (58K):
[in this window]
[in a new window]
|
Figure 9. There is difficulty in distinguishing the talar dome (TD) from the distal tibial epiphyseal edge on the B-scan (a), however a coronal reformat of the data (b) helps to identify the superior margin of the talar dome as illustrated by landmark X on both the reformatted and original images. On image (b) the width of the medial malleolus (MM) is just 3.7 mm. (c) An annotation of the reformatted data.
|
|
View larger version (64K):
[in this window]
[in a new window]
|
Figure 10. (a) The original B-scan. (b) There is clearer definition of the cartilaginous border of the navicular having displayed the maximum intensity projection from a 1 mm stack of reformatted images either side of the original B-scan.
|
|
From the hind-foot data set only the superficial surface of the diametaphyseal region of the ossified tibia and fibula was displayed as a highly reflective semi-circle and segmentation of their complete structure was in part estimated (Figure 11a,b). The epiphyses of the tibia and fibula are non-ossified allowing their position at the ankle joint to be directly outlined rather than estimated (Figure 12). The talar body does contain a central ossified nucleus at this age resulting in partial estimation of its lateral surface (Figure 13), however the distal part of the talar neck and the head are cartilaginous. The position of the calcaneum and cuboid could not be identified from this particular approach due to their relatively posterior and lateral position with respect to the medially placed transducer, and therefore are not included in the final reconstruction.
View larger version (78K):
[in this window]
[in a new window]
|
Figure 11. (a) Illustration of the ossified distal shaft of the tibia and fibula in transverse section from original images. Due to the high reflectivity of ossified bone, segmentation was in part estimated (b).
|
|
View larger version (128K):
[in this window]
[in a new window]
|
Figure 12. An original B-scan from the data set illustrates the low reflectivity of the tibial (te) and fibular epiphyses (fe).
|
|
View larger version (45K):
[in this window]
[in a new window]
|
Figure 13. (a) A 1 mm maximum intensity projection through the data set in the same plane as the original B-scans, illustrates the talus with respect to the medial malleolus and navicular. Its ossified central nucleus prohibits accurate segmentation of its lateral border with this orientation of the transducer. (b) Illustration of the annotated anatomy.
|
|
From the forefoot data set, segmentation of the talar head, navicular and cuneiforms could be performed with the aid of reformatting techniques as described. The proximal epiphysis of the first metatarsal was also visible from this approach. On completion of the manual segmentation of both the forefoot and hind foot data sets, surface rendered outlines of the segmented bones in relation to one another is shown in three dimensions (Figures 14 and 15). No attempt is made to differentiate between ossified and non-ossified bone, therefore the surface rendered outline resembles a fused skeleton.
View larger version (36K):
[in this window]
[in a new window]
|
Figure 14. Two projections of a surface rendered left hindfoot with the distal shafts of the tibia and fibula, viewed from (a) the front and (b) from above showing the talonavicular alignment.
|
|
On average there were four satisfactory data sets per foot examined. The surface rendered reconstructions for each foot were subjectively reproducible. A quantitative assessment of reproducibility was beyond the scope of this feasibility study.
|
Discussion
|
|---|
A thorough knowledge of normal anatomy is essential when trying to understand anatomical deformities, particularly when these deformities are complex and multiplanar such as in CTE. The diagnosis of neonatal foot deformities is generally clinical. Currently plain radiography is the only imaging method occasionally used in the assessment and follow-up of such babies, but is hampered by the lack of ossification for much of the first year of life. It is well recognized that ossification centres may originate eccentrically within their cartilaginous anlages [14] and this eccentricity may be more pronounced in CTE [15]. Therefore whilst plain radiographs are still used in the assessment and follow up of CTE, they offer only limited information about the precise location of the tarsal structures with respect to one another, especially during the first year of life. In order for such imaging to be useful, reference measurements and angles are required from a normal population and such data are available [16] regarding plain radiography and CTE. MRI and ultrasound have been used as research tools to define the anatomy of the small bones [2–4]. Sedation is frequently required for a neonatal MRI examination. MRI is a relatively lengthy procedure compared with the data acquisition stage described in this study which takes just a few seconds. Aurrell et al [12] describe a method for examining the normal and abnormal foot with standard 2D ultrasound, using medial, lateral and dorsal scanning planes. Schlesinger et al describe how 2D ultrasound was used to demonstrate dorsal talonavicular dislocation in an infant with congenital vertical talus [19]. The advantage of a 3D US data set is that from one sweep of the transducer, not only can the 3 different 2D planes described by Aurrell be produced, but 3D definition of tarsal orientation is possible.
This pilot study illustrates that it is possible to obtain a 3D image of selected structures of the neonatal foot using 3D ultrasound. Owing to central ossification of the talus, calcaneum and cuboid, resulting in acoustic shadowing, a full 3D outline of the skeletal anatomy of the neonatal foot using a single 3D ultrasound data set is not possible. However by using several different approaches, separate aspects of the foot may be examined, depending on the area of clinical interest. For example, 3D ultrasound of the neonatal foot may have an important clinical application in congenital talipes equinovarus, where there is medial subluxation of the navicular and the talar neck is medially deviated with respect to the talar body [15, 17, 18]. These deformities could be displayed with 3D US using the medial approach of the hind-foot. For the more rare hind foot deformities such as calcaneo-valgus, a lateral approach of the transducer might provide the necessary information. In theory, multiple sweeps of the transducer over the foot using different projections, could produce overlapping data sets. These data sets could be merged providing data that would cover the whole foot. However, unless the foot was held absolutely rigid with the child totally immobile, there would be errors caused by foot movement between acquisitions and variations in the slight transducer pressure applied to the foot during an acquisition, to allow accurate merging of the data.
A disadvantage of this method of examining the neonatal foot is that it requires the infant to be completely immobile. In this study only babies that were asleep or about to be fed were studied. Should such a procedure be implemented in clinical practice, only a light sedative would be required due to the short duration of the examination (5–8 s per data set). At less than 6 weeks of age most infants spend much of the day asleep and in our experience, an additional sedative would frequently not be necessary.
Estimating the entire outline of a partially ossified structure will lead to unavoidable inaccuracies due to acoustic shadowing (Figure 13). In practice, the articular surface of important structures are non-ossified (e.g. the distal pole of the talus, navicular, fibular and tibial epiphyses) and this limitation should not result in any errors in defining structural relationships. Three important clinical relationships may be defined by segmenting around clearly visible structures. These comprise (a) the position of the talus within the ankle mortice, (b) the angle the talar neck makes with the body, and (c) the position of the navicular in relation to the medial malleolus and anterior pole of the talar head.
An accompanying MRI study to validate these findings would have been ideal, but it was considered ethically unacceptable to subject a healthy neonate to sedation or to a general anaesthetic that may be required for such a procedure.
The surface rendered images produced in this study resemble a fused skeleton. Separate segmentation of the tibial and fibular epiphyses would be possible but the segmentation of a minute structure such as the fibular epiphysis (approximately 3 mm in width) would be inaccurate, and it would be difficult to replicate the exact position of the end plates given the orientation of the transducer. Rather than introduce a further set of potential errors, the tibia and fibula were segmented to include their epiphyses as if they were fused.
The soft tissue component of a foot deformity is also important to the surgeon. No attempt, however, has been made to reconstruct the soft tissues of the foot. Although it may be technically possible to outline tendons and ligaments, these structures are considerably smaller than their cartilaginous or bony counterparts, increasing the inaccuracies of reconstruction. MRI may provide the necessary soft tissue detail prior to a tendon transfer, although this surgical intervention tends not to be the primary procedure.
The 3D US system described has been extensively validated in the laboratory and demonstrates a point location accuracy of 0.5 mm [9]. If 3D US of the neonatal foot is to be used in clinical practice, a robust repeatable method of examining the normal foot is required with supporting clinical validation. To our knowledge there are no published data describing 3D normal neonatal foot anatomy using ultrasound or other imaging techniques. If such a method can be produced, 3D ultrasound of the deformed foot would be feasible and may offer the surgeon a greater understanding of the abnormal bony anatomy prior to any surgical intervention, or the ability to monitor any improvement during a period of conservative non-operative management.
Received for publication September 1, 2004.
Revision received November 6, 2004.
Accepted for publication January 26, 2005.
|
References
|
|---|
- Simons GW. A standardized method for the radiographic evaluation of clubfeet. Clin Orthop Related Res 1978;135:107–18.
- Downey DJ, Drennan JC, Garcia JF. Magnetic resonance image findings in congenital talipes equinovarus. J Pediatr Orthop 1992;12:224–8.[Medline]
- Grayhack JJ, Zawin JK, Shore RM, Trombino LJ, Posnanski AK, Carroll NC. Assessment of calcaneocuboid joint deformity by magnetic resonance imaging in talipes equinovarus. J Pediatr Orthop B 1995;4:36–8.[Medline]
- Kamegaya M, Shinohara Y, Kuniyoshi, Moriya H. MRI study of talonavicular alignment in clubfoot. J Bone Joint Surg (Br) 2001;83–B:726–30.
- Tolat V, Boothroyd A, Carty H, Klenerman L. Ultrasound: a helpful guide in the treatment of congenital talipes equinovarus. J Pediatr Orthop B 1995;4:65–70.[Medline]
- Napiontek M. Intraoperative ultrasound for evaluation of reduction in congenital talipes equinovarus. J Pediatr Orthop B 1995;4:55–7.[Medline]
- Hamel J, Becker J. Sonographic assessment of clubfoot deformity in young children. J Pediatr Orthop B 1996;5:279–86.[Medline]
- Cash CJC, Berman LH, Treece GM, Gee AH, Prager RW. Two and three dimensional ultrasound in the development of a needle-free injection system. Br J Radiol 2004;77:236–42.[Abstract/Free Full Text]
- Treece GM, Gee AH, Prager RW, Cash CJC, Berman L. High definition freehand 3D ultrasound. Ultrasound Med Biol 2003;29:529–46.[CrossRef][Medline]
- Treece GM, Prager RW, Gee AH, Berman L. Correction of probe pressure artefacts in freehand 3D ultrasound. Med Imaging Anal 2002;6:199–214.
- Treece GM, Prager RW, Gee AH, Berman L. Surface interpolation from sparse cross-sections using region correspondence. IEEE Trans Med Imaging 2000;19:1106–14.[CrossRef][Medline]
- Aurrell Y, Johansson A, Hansson G, Wallander H, Jonsson K. Ultrasound anatomy in the normal neonatal and infant foot: an anatomic introduction to ultrasound assessment of foot deformities. Eur Radiol 2002;12:2306–12.[Medline]
- Gee AH, Prager RW, Treece GM, Berman L. Narrow-band volume rendering for freehand 3D ultrasound. Comput Graph 2002;26:463–76.[CrossRef]
- Hubbard AM, Meyer JS, Davidson RS, Mahboubi S, Harty MP. Relationship between the ossification center and cartilaginous anlage in the normal hindfoot in children: study with MR imaging. AJR Am J Roentgenol 1993;161:849–53.[Abstract/Free Full Text]
- Shapiro F, Glimcher MJ. Gross and histological abnormalities of the talus in congenital club foot. J Bone Joint Surg 1979;61A:522–30.
- Vanderwilde R, Staheli LT, Chew DE, Malagon V. Measurements on radiographs of the foot in normal infants and children. J Bone Joint Surg 1988;70-A:407–15.
- Ippolito E. Update on pathologic anatomy of clubfoot. J Pediatr Orthop B 1995;4:17–24.[Medline]
- Irani RN, Sherman MS. The pathological anatomy of club foot. J Bone Joint Surg 1963;45A:45–52.
- Schlesinger AE, Deeney VFX, Caskey PF. Sonography of the nonossified tarsal navicular cartilage in an infant with congenital vertical talus. Pediatr Radiol 1989;20:134–5.[CrossRef][Medline]
Top
Abstract
Introduction
