A video digitizer for analysis of trunk deformity in scoliosis

Communications A video digitizer for analysis of trunk deformity in scoliosis S.H. Slupsky*,

N.G. Durdle’,

V.J. Raso $, D.L. Hill’ and A.E. Peterson*’

*Applications Engineering, Alberta Microelectronic Centre; ‘Department of Electrical Engineering, University of Alberta; ‘Orthopaedic Biomechanics, Glenrose Rehabilitation of Civil Engineering, University of Alberta Hospital, Edmonton; and *‘Department Received August 1990, accepted July 1991

ABSTRACT Scoliosis is a deformity characterized by lateral curvature of the spine and accompanied by axial rotation of the vertebrae; it ofien causes varying degrees of trunk deformity. Research has indicated that topographic techniques can be used to describe the disorder and monitor its progression. A video image acquisition system has been designed which reduces the time required to quantlfi topographic details of the trunk and aids in the diagnosis, monitoring and research of scoliosis. This system integrates the capability of large, expensive grey-scale image acquisition equipment into a small, low-cost diagnostic imaging tool using current technologies and design techniques. The video digitizer accepts a standard NTSC monochromatic video signal as input and the unit is connected to a computer via an IEEE-488 bus from which the ditigivr is controlled. The digitizer samples the video signal in real time using a high-speedfih converter controlkd by an application-speciJc integrated circuit; the digital sampixs are stored in memory until the host computer requests that the information be transferred. Keywords:

Digitizer, video, scoliosis

INTRODUCTION Scoliosis is an abnormal lateral curvature of the spine combined with vertebral rotation. The deformity may arise for a number of reasons: muscle imbalance, neurological deficits, growth disturbances (congenital or due to radiation therapy) or from the development of a spinal tumour; but the most common type, idiopathic scoliosis, develops in children for unknown reasons, and is readily ap arent during periods of rapid growth. Consequent Py, the adolesassociated with the cent growth spurt is frequent1 discovery of scoliosis. These c Kildren are normal in virtually every respect except for the presence of this curvature; there are no obvious characteristics which can be used to screen children at risk. Yet early detection is essential both for effective treatment and for understanding the cause(s) and development of this deformity. One of the poorly understood facets of idiopathic scoliosis is the relation between spine deformity and trunk deformity. There should be a direct relation between surface topography and known deformities of the spine and ribs, but in reality, similar s inal curves ap ear to cause quite different trunk B eforissue because most mities. Tl! is is an important children (or their parents) seek treatment because they are unhappy with the cosmetic appearance of Correspondence and reprint requests to: Dr Nelson G. Durdle, Dept. of Electrical Engineering, Civil/Electrical Engineering Bldg, University of Alberta, Edmonton, Alberta, Canada, T6G 2G7 0 1992 Butterworth-Heinemann 0141-5&25/92/010069-04

their back. For this same reason, the physician may recommend certain non-o erative treatments based on the perceived effect tKese may have on trunk deformity, even though the treatments may not significantly alter internal spinal alignment. Traditionally, children with scoliosis are monitored by X-ray pictures. This is necessary for those with scoliosis who are at significant risk of progression. Unfortunately X-rays do not record the surface shape well and cannot be used, due to ionizing radiation, to screen and follow large groups for possible presence of subclinical spinal deformity. and progression Radiation risks associated with this procedure are of particular importance since the majority of those screened will be found to have no spinal deformity ‘. For these reasons there has been considerable work done to develop alternative methods for documenting trunk deformity of children with idiopathic scoliosis. Some researchers 2,3,4 have been unable to establish a link between spinal deformity and trunk asymmet wi 3: by analysing trunk deformity in children scoliosis. However, Turner and Smith” using ‘ISIS’ claim a good correlation. Takasaki” pioneered the application of moire fringe topography, the most commonly used technique to represent three-dimensional topography in in two dimensions, to the study of trunk deformi children. This was essentially a qualitative tool wx ich provided a quick assessment of trunk symmetry and was used for screening. Roger7 has shown that the method could be analysed to provide coordinate data of trunk deformity. In this method a line pattern is

for BES J. Biomed. Eng. 1992, Vol. 14, January

69

Video digitkrfir

scoliosisstudies: S.H. Supsky et al.

pro’ected on to the back of a subject and combined wit h a reference grating to generate moire fringes. The analysis of moire fringes is limited by the labour and time required to digitize the fringe patterns manually. In a typical clinical system, the errors involved are theoretically of the order of f 2 mm of surface height, increasing to almost 5 mm if only minimal subject position support is provideds. In addition, traditional moire techniques do not lend themselves to digital processing and the use of numerical techniques, which might reduce measurement error and improve reliability. An alternative video-based system5 available for the assessment of trunk deformity provides profiles at discrete levels of the trunk. Known as ISIS (Integrated Shape Investigation System), this technique is limited because of the relatively long time needed to image each scan, low spatial resolution and poor repeatability. Also, there is no interaction capability with this system. Scoliosis is a complex three-dimensional deformity, and past research efforts have shown that a single view is of limited value. Drerup’ and Heirholzer’ have developed video techniques based on specialized hardware which matched the camera and ‘frame grabber’ resolution. This matching is essential if data are to be interpolated at the subpixel level. Lack of significant advances in this area highlight the need for improved methods for analysis of back topography. The clinical objective of the work described is to provide a topographical map of the back and interactive graphics tools which will permit the clinician to examine the trunk at any relevant angle. A system which has been specifically designed to acquire video images of the trunk is described; it provides a fast, accurate, reliable and economical method of digitizing images and transferring them to a computer for further processing. To accomplish this, a system has been developed to capture, in real time, up to four frames of a video image with a resolution of 5 12 x 484 pixels. The digitized image is then transferred to a host computer through an IEEE-488 instrumentation bus. IEEE-488 interface provides flexibility and independence of computer systems since IEEE488 interfaces are readily available for IBM PC, Macintosh and most engineering workstations. The facility to capture up to four frames in real time provides for frame averaging to minimize the effects of patient movement. The system is controlled by a Motorola MC6809 microcomputer using a ROMbased program. This program can be changed easily to accommodate particular application requirements. The use of an application-specific integrated circuit reduces cost and minimizes the board area required to implement the system. It also reduces the noise problems associated with operating a system at 40 MHz. THE VIDEO ACQUISITION

SYSTEM

The video acquisition system has both hardware and software components. The hardware system samples a National Television Standards Committee (NTSC) type video image’a in real time, with sufficient pixel resolution and pixel depth to retain ali information contained in the original video signal. The system

70 J. Biomed. Eng. 1992, Vol. 14,January

Video

VMA * Write

SYOC Timing Analogtie

Read -

*

Video

Video

Data

Figure 1 Block diagram of video digitizer

interfaces to other computers 488 instrumentation bus.

via a high-speed

IEEE-

Hardware component of the video digitizer The video digitizer (block diagram shown in Figure 7) uses a MC 6809 microprocessor to control the hardware. The main tasks of the microprocessor are system control and coordination of data transfer. The microprocessor programs a video Direct Memory Access (DMA) circuit and controls communication. This dedicated DMA circuit was designed to accommodate the high data rate associated with the real-time sampling of the video signal. The video memory subsystem comprises 2.56 kilobytes of 1OOns CMOS static RAM and an address decoder. To accommodate the high-speed data transfer (10 MHz), it was necessary to use high-speed RAM and address decoding. Static RAM was chosen for its ease of use and cost, which is comparable to highspeed dynamic RAM. The analogue video signal processor conditions the incoming video signal and generates synchronization and timing signals. The signal conditioning provides a low-pass filtered video signal to eliminate aliasing effects. The signal conditioning circuit also extracts s nchronization pulses which are then routed to and d ecoded by a Motorola 6840 Programmable Timer Module (PTM). The video sampling interface is shown in Figure 2. This circuitry consists of an A/D converter, reference voltage circuitry, and a eripheral A/D interface module that controls the B ow of data from the A/D converter to the video memory. A video frame is sampled in real time at 10 MHz to generate a 512 X 484 pixel image. A flash A/D converter is used to attain the required sampling rate. To account for varying lighting conditions, low contrast images, and spatially variant contrast images due to skin reflectivity, a minimum of 256 quantization levels (8 bits) were required to represent the video signal adequately ’ I. The Peripheral A/D Interface Module (PADIM) controls the flow of data from the high-speed A/D converter to video memory using direct memory access. The PADIM is implemented as a 2500 gate application-specific integrated circuit using two micron gate array technology (device designed using

Video dipitizerfor scoliosisstudies: S. H. Slupsky et al. ASK VMA

Data

( Peripheral AID interface module

Timing *

Write

5 ~

Read L. Address

Start

Data

Enable

* w

Video

w A/D Converter Reference voltage circuitry

B

w

‘

Figure 2

Block diagram

of video sample interface

the LDS design tools from LSI Logic Corporation of Milpitas, CA and fabricated at the Alberta Microelectronic Centre, Edmonton, AB.) A gate array was chosen to reduce circuit board complexity and cost, and to improve reliability of the system. The PADIM provides an interface between the microprocessor and the video memory through a 32K memory block in the microprocessor address space. The device can be programmed with the appropriate page address to access all video memory. The video memory address space of the PADIM was designed to accommodate up to one megabyte of RAM. Another feature of the interface module is the ability to program a delay register that accommodates the pipe-lining effect of some A/D converters. When the first sample is taken the encoded binary number may not appear immediately but rather at some fixed number of samples later at the output of the AD converter. The interface module can compensate for a delay of 1 to 8 sample periods. Microprocessor

software

The structure of the ROM-based software is shown in Figure 3. The software first initializes the microproces-

Figure 3

Software diagram

the video system and the IEEE-488 communications interface. The microcomputer then polls the IEEE-488 interface for pending messages from a host computer. If the host command indicates a digitize video field action, the software sets up the memory address for the video frame, digitizes the video frame and sends a response to the host computer indicating the completion of the frame acquisition. If the host command indicates an up-load action, the computer transfers the digitized video information to the host computer system. to the video frame, required Synchronization before sampling of the video image can proceed, is accomplished by detecting the end of the second field of a video frame. First, the program waits for a vertical by a horizontal synchronization pulse followed synchronization pulse. After this, the program waits for a video equalization pulse. When an equalization pulse is detected, the module measures the pulse width of the preceding video line (the last visible video line). The width of this line indicates the current field of the video frame and consequently synchronization. Once frame synchronization has been established, the module digitizes both video fields using the Digitize Field Module. The Digitize Field Module sets up the base address of the first video line. The blanking delay circuit is then activated, which in turn provides the necessary start signal to the PADIM. The software simultaneously reprograms the PADIM with line addresses while the memory transfers are taking place. This continues until the program detects that the required number of video lines for a field (242 by default) have been sampled. When all the video lines of the current field have been digitized, the module disables the blanking delay circuit. SOT,

Address

CLINICAL

APPLICATION

This board will be used in the clinical assessment of trunk deformity in children with idiopathic scoliosis. Measurement of trunk deformity based on moire fringe patterns was time consuming because of the special photographic processing involved; it often took weeks from the initial photograph to the time when the negatives were available for analysis. The and required analysis step was time consuming manual digitizing of fringe data. In a growing child the length of time this takes can often make the results irrelevant with regard to immediate treatment decisions. The technology we have described will permit trunk deformity measurements to be available immediately to the clinician. In a single review the clinician will have the patient, the X-ray describing the internal deformity and objective measures of the trunk deformity. A preliminary approach to the determination of trunk topography is to analyse a series of grid lines projected on to the back. The deformation of these grid lines provides a measure of the trunk topography. While profiles may be obtained directly along the grid-line, it would be necessary to interpolate between lines to map the entire trunk. Illustrated in Figure 4 is an example of a 1.5year-old girl with idiopathic scoliosis and a right thoracic curve of

J. Biomed.

Eng.

1992, Vol.

14, Janu~y

71

forscoliosisstudies: S.H.

Video digitizer

Slupsky et al.

no advanced image processing methods. This example illustrates the potential and the feasibility of the system described.

DISCUSSION Since current techniques to quantify scoliosis expose the patient to harmful radiation, new topographic techniques, involving the use of safe, non-invasive, non-radiating light and digital analysis of the topographic images, can contribute significantly to the treatment and the following of children with scoliosis. The video digitizer’s performance and circuit utting more complexity could be improved by functions on the ASIC. Functional bloc Ks, such as the digital portion of the Analogue Video Signal Processor (MC6840 PTM and associated circuitry) and a hardware filter, could have been designed into the integrated circuit. This would further decrease the chip count of the digitizer and increase reliability. With the addition of a hardware digital filter, time required to quantify results can be reduced. At present, topographic analysis must be performed on another computer system which lengthens the analysis time, thus limiting the use of the digitizer. It would be desirable to design the hardware to perform analysis of the topographic image in real time; however, further research must be conducted into the image processing techniques required to produce reliable trunk measurements.

REFERENCES Figure 4

Video image of a patient with scoliosis

47”. A series of lines is projected on to her back from a standard 35mm slide projector. A simple analysis of this image provides the profiles shown in Figure 5. In a normal child these profiles would each be symmetrical. There is a slight asymmetry in profile A while profile B is more asymmetrical. Profile B also demonstrates the degree of trunk rotation. This important clinical information was gleaned from the image quickly, using a simple technique and virtually

loor

-100

t

1

I

I

100

200

300

I

400

Horizontal fringe

Figure 5 Typical trunk profiles. -

72

fringe A; ----

J. Biomed. Eng. 1992, Vol. 14, January

fringe B

1. Nottage WM, Waugh TR, McMaster WC. Radiation exposure during scoliosis screening radiography. Spine 1981; 6: 456-9. Sahlstrand T. The clinical value of moire topography in the management of scoliosis. Spine 1986; 11: 409-17. Stokes IAF, Shuma-Hartswick D, Moreland MS. Spine and back-shape changes in scoliosis. Acta Orthop &and 1988; 59: 128-33. Drerup B, Hierholzer E. First experiences with clinical applications of video rasterstereography. In: Alberti A, ed. Proc Vlth Int Symp Surface and Spinal Deformity, Universidade Nova De Lisboa, 19-20 Sept., 1990. .5. Turner-Smith AR, Harris JD. ISIS - An automated shape measurement and analysis system. In: Harris JD, ed. Surface Topography and Spinal Deformity. Stuttgart and New York: Gustav Fischer, 1986. Takasaki H. Moire topography. Applied Optics 1973: 12: 845-50. Roger R. Shape Measurement with Moire Topography [PhD Thesis]. Exeter College Oxford, 1980. Wegner J. The Measurement of Scoliosis using Moire Topography [MSc Thesis]. Edmonton, Canada: University of Alberta, 1985. 9. Hierholzer E. Improved methods of image processing in video rasterstereography. In: Alberti A, ed. Proc VZthInt Symp Surface and Spinal Deformity, Universidade Nova De Lisboa, 19-20 Sept., 1990. 10. Kivar MS, Kaufman M. Television Electronics. Delmar Publishers, 1983. 11. Slupsk SH. A Video Digitizer for Scoliosis Studies [MSc Thesis fy . Edmonton, Canada: Dept. of Electical Engineering, University of Alberta, 1988.