30-03-2011, 09:20 AM
presented by
AMEER BASHA
SANJEEVIAH
[attachment=11287]
ABSTRACT
Medical imaging is a field which researches and develops tools and technology to acquire, manipulate and archive digital images which are used by the dimensional function , f, that takes an input two spatial coordinates x and y and returns a value f(x, y). The value f(x, y) is a gray level of the image at that point. The gray level is also called the intensity. Digital images are a discretized partition of the spatial images into small cells which are referred to as pixels – picture elements. Digital image processing is a field for processing digital images using a digital computer. Processing of digital images include operations involving digital images such as acquisition, storage, retrieval, translation, compression, etc.
Quantifying disease progression of patients with early stage Rheumatoid Arthritis (RA) presents special challenges. Establishing a robust and reliable method that combines the ACR criteria with bone and soft-tissue measurement techniques, would make possible the diagnosis of early RA and/or the monitoring of the progress of the disease. In this paper and automated, reliable and robust system that combines the ACR criteria with radiographic absorptiometry based bone and soft tissue density measurement techniques is presented. The system is comprised of an image digitization component and automated image analysis component. Radiographs of the hands and the calibration wedges are acquired and digitized following a standardized procedure. The image analysis system segments the relevant joints into soft-tissue and bone regions and computes density values of each of these regions relative to the density of the reference wedges. Each of the joints is also scored by trained radiologists using the well established ACR criteria. The results of this work indicate that use of standardized imaging procedures and robust image analysis techniques can significantly improve the reliability of quantitative measurements for rheumatoid arthritis assessment. Furthermore, the methodology has the potential to be clinically used in assessing disease condition of early stage RA subjects.
Keywords: Automated hand image analysis, Hand image segmentation, Radiographic absorptiometry, Rheumatoid arthritis
1. INTRODUCTION
Conventional examination of the hand radiographs is well established as a diagnostic as well as an outcome measuree in Rheumatoid Arthritis (RA). It is readily available and has been correlated with measures of disease activity and function. X-ray changes are, however, historical rather than predictive, and there is significant observer variation in quantifying erosive changes. The earliest radiographic changes seen in the hand are soft-tissue swelling symmetrically around the joints involved, juxta-articular osteoporosis and erosion of the ‘bare’ areas of bone (i.e. areas lacking articular cartilage).These changes help to confirm the presence of an inflammatory process.
The presence of early soft-tissue swelling is easily recognized on plain radiographs but not readily quantified. Although the presence of early osteoporosis is recognized in the affected hand, a mild osteoporosis may be extremely subtle to the eyes. The recognition of the changes in soft-tissue and bone density is subjective and is known to vary from assessor to assessor. Therefore, attention has been focused on the more objective erosion and joint narrowing assessment. Use of magnetic resonance (MR) technique has been shown to sensitively detect early local edema and inflammation prior to a positive finding on plain film radiographs. However, MR is an expensive examination and may not be used as a routine technique.
Presently, radiographs of the hands and wrists are employed to assess disease progression. The parameters used to determine progression are the changes in erosions and joint-space narrowing observed on the radiographs. There are some problems with both of these parameters. First, both erosion and joint narrowing are not the earliest changes in RA and further they may be substantially irreversible. Second, these two changes may occur independently of each other. Third, there is a tremendous variability in erosive disease: some patients never develop erosions; some go into spontaneous remission of their erosive disease; and for some, the progression is relentless. Fourth, joint-space or cartilage loss may be caused by either the disease itself or by mechanical stress. Present scoring methods require that any degree of joint-space loss be recorded as a progressive change due to RA.
Quantitative techniques currently available may provide a new approach in monitoring disease progression in patients with RA. Adoption of these techniques may have implications for the management of patients with RA and for possible detection of the disease at an early stage.
1.1. Hand bone densitometry
Considerable advances have been made over the past two decades in developing radiological techniques for assessing bone density. However; all of these techniques have been utilized on aging-related osteoporosis, a pathological change involving general bone mineral reduction. Owing to the wide availability of DXA, recently published research describes the use of Bone Mineral Density (BMD) measurements in the hands of patients with chronic RA. Most published observations on RA have examined BMD changes, focusing on only the general bone loss around the joints. Quantification of the difference of bone loss between the juxta-articular bone and the shaft of tubular bones in the hands could be a sensitive index for quantitative analysis of RA patients. Hand BMD measurements offer an observer independent and reproducible means of quantifying the cumulative effects of local and systemic inflammation. The technique could be of use in the assessment of patients with early RA, in whom conventional measures of disease are not helpful until disease is (irreversibly) more advanced.
1.2. Hand radiographic absorptiometry
In conventional Radiographic Absorptiometry, radiographs of the hand are acquired with reference wedges placed on the films. The films are and subsequently analyzed using an optical densitometer. The resulting density values computed by the densitometer are calibrated relative to that of the reference wedge and are expressed in arbitrary units.
Recent improvements in hardware and software available for digital image processing have led to the quantitative assessment of radiological abnormalities in diagnostic radiology. Such improvements have also enabled introduction of several radiographic absorptiometry techniques. One such technique uses centralized analysis of hand radiographs and averages the BMD of the second to fourth middle phalanges. Another technique developed in Japan uses the diaphysis of the second metacarpal to determine BMD. A third technique developed in Europe measures the diaphysis and proximal metaphysics of the second middle phalanx. Based on published short-term precision errors, computer-assisted Radiographic Absorptiometry appears to be suitable for the measurements of the BMD of phalanges and metacarpals, and is used in several hundred centers worldwide.
In this work we present preliminary results of an ongoing research work aimed at developing an automated radiographic absorptiometry system for the assessment and monitoring of both BMD and soft tissue swelling in early stage RA. This paper focuses on the reproducibility and accuracy of the methodology being developed. The paper is organized as follows: the next section provides an overview of the image acquisition procedure. In section 3 the image analysis algorithms used in this work are presented. In section 4 we present results obtained by analyzing the data collected in a small reproducibility study involving 10 normal subjects.
2. IMAGE ACQUISITION
One key factor influencing the outcome of any radiographic absorptiometry technique is the standardization of the image acquisition technique. Variability in acquisition parameters can significantly affect the measured values. In order to carry out this work, a standard image acquisition protocol was defined. This protocol has been successfully used in earlier large scale multinational phase 3 clinical trials for Rheuatoid Arthrtis related drugs. Radiographs of the left and right hands are taken one
at a time. Templates were developed to guide the positioning of the hand with respect to the center of the x-ray beam. The X-ray beam was centered between the 2nd and 3rd metacarpo-phalangeal joints and angled at 90° to the film surface. This results in a tangential image of the joints. Improper beam centering generally results in overlapping joint margins. The X-ray exposure parameters were maintained constant for all subjects. All normal subjects were imaged at the same clinic at UCSF. In addition to providing a template for hand positioning, two sets of calibration wedges were also provided to the clinic. Each set of wedges consisted of one Acrylic wedge, for soft tissue and one Aluminum wedge for bone tissue. These wedges were custom designed for the purposes of this research work.
3.Enhancement Image and Restoration
The image at the left of Figure 1 has been corrupted by noise during the digitization process. The 'clean' image at the right of Figure 1 was obtained by applying a median filter to the image.
An image with poor contrast, such as the one at the left of Figure 2, can be improved by adjusting the image histogram to produce the image shown at the right of Figure 2.
3.IMAGEANALYSIS
One of the major difficulties in analyzing hand radiograph images is the high level of noise present in the images. Additionally the trabecular texture of the hand in the vicinity of the joints increases the noise in edge maps of this regions. Use of non-standard acquisition protocol can add additional challenges at it can result in further degradation of image quality. This last challenge is minimized in this work, as a standard image acquisition protocol is followed. Given a particular application varying degrees of accuracy in anatomy segmentation can be considered acceptable. For instance in detecting joint-space narrowing there is a need for accurate and reliable determination of the joint-space of any finger and the bone edges in this region. However, accurate delineation of the bone edges in the vicinity of the joint is not as relevant. Depending upon the application there can be additional constraints on performance issues as well. In an application for which off-line processing of the data is acceptable, more sophisticated algorithms can be employed. This particular application requires that the overall process be fast, accurate and reproducible enough for on-line processing. Accurate estimation of the bone edges in the middle shaft and in the joint vicinity of high relevance in this work. This is primarily because the disease progression follows different patterns in the joint area as compared to the middle phalange area. Also, the manifestations of the disease symptoms in its early stage have different effects on soft-tissue and bone as well, which require reliable segmentation of these two types of tissue at different time points. The algorithm for hand segmentation can be outlined into the following main stages:
• Hand outline delineation
• Joint identification
• Bone outline delineation
• Segmentation of soft tissue and bone
The first stage of the algorithm has been well studied in the literature and is not described here. The second stage can be more challenging, especially when dealing with hands of patients in advanced stage of disease progression. As this system will be applied to a patient population that is in their early disease stage it is expected that the joints will be well defined. The system provided to the radiologists allows them to adjust the location of the automatically identified joints. Results presented in this work were obtained by having the radiologists place control points to identify the joints, rather than having them automatically computed by the system.
3.1. Control point placement
A simple user interface was provided to enable placement of control points on the joints. This was primarily done to investigate the sensitivity of the system to the initial control point positioning, which in an automated system would invariably be the same for the same image. The user placed 16 control points on each image. These joints are show in Figure 4. In addition to placing the control points for the joints, the control points for the two wedges are also placed by the radiologists. For each wedge, six control points are placed with four at the corners and two in the middle. Once all the control points are placed, the remaining steps of the generalized algorithm stated above are carried out autonomously. The middle phalange or cortex control points are computed automatically and are located at the middle of straight line connecting the two joints, one above and one below the middle phalange. The diameters of the circular regions of interest placed around each joint are computed proportionally to the length
AMEER BASHA
SANJEEVIAH
[attachment=11287]
ABSTRACT
Medical imaging is a field which researches and develops tools and technology to acquire, manipulate and archive digital images which are used by the dimensional function , f, that takes an input two spatial coordinates x and y and returns a value f(x, y). The value f(x, y) is a gray level of the image at that point. The gray level is also called the intensity. Digital images are a discretized partition of the spatial images into small cells which are referred to as pixels – picture elements. Digital image processing is a field for processing digital images using a digital computer. Processing of digital images include operations involving digital images such as acquisition, storage, retrieval, translation, compression, etc.
Quantifying disease progression of patients with early stage Rheumatoid Arthritis (RA) presents special challenges. Establishing a robust and reliable method that combines the ACR criteria with bone and soft-tissue measurement techniques, would make possible the diagnosis of early RA and/or the monitoring of the progress of the disease. In this paper and automated, reliable and robust system that combines the ACR criteria with radiographic absorptiometry based bone and soft tissue density measurement techniques is presented. The system is comprised of an image digitization component and automated image analysis component. Radiographs of the hands and the calibration wedges are acquired and digitized following a standardized procedure. The image analysis system segments the relevant joints into soft-tissue and bone regions and computes density values of each of these regions relative to the density of the reference wedges. Each of the joints is also scored by trained radiologists using the well established ACR criteria. The results of this work indicate that use of standardized imaging procedures and robust image analysis techniques can significantly improve the reliability of quantitative measurements for rheumatoid arthritis assessment. Furthermore, the methodology has the potential to be clinically used in assessing disease condition of early stage RA subjects.
Keywords: Automated hand image analysis, Hand image segmentation, Radiographic absorptiometry, Rheumatoid arthritis
1. INTRODUCTION
Conventional examination of the hand radiographs is well established as a diagnostic as well as an outcome measuree in Rheumatoid Arthritis (RA). It is readily available and has been correlated with measures of disease activity and function. X-ray changes are, however, historical rather than predictive, and there is significant observer variation in quantifying erosive changes. The earliest radiographic changes seen in the hand are soft-tissue swelling symmetrically around the joints involved, juxta-articular osteoporosis and erosion of the ‘bare’ areas of bone (i.e. areas lacking articular cartilage).These changes help to confirm the presence of an inflammatory process.
The presence of early soft-tissue swelling is easily recognized on plain radiographs but not readily quantified. Although the presence of early osteoporosis is recognized in the affected hand, a mild osteoporosis may be extremely subtle to the eyes. The recognition of the changes in soft-tissue and bone density is subjective and is known to vary from assessor to assessor. Therefore, attention has been focused on the more objective erosion and joint narrowing assessment. Use of magnetic resonance (MR) technique has been shown to sensitively detect early local edema and inflammation prior to a positive finding on plain film radiographs. However, MR is an expensive examination and may not be used as a routine technique.
Presently, radiographs of the hands and wrists are employed to assess disease progression. The parameters used to determine progression are the changes in erosions and joint-space narrowing observed on the radiographs. There are some problems with both of these parameters. First, both erosion and joint narrowing are not the earliest changes in RA and further they may be substantially irreversible. Second, these two changes may occur independently of each other. Third, there is a tremendous variability in erosive disease: some patients never develop erosions; some go into spontaneous remission of their erosive disease; and for some, the progression is relentless. Fourth, joint-space or cartilage loss may be caused by either the disease itself or by mechanical stress. Present scoring methods require that any degree of joint-space loss be recorded as a progressive change due to RA.
Quantitative techniques currently available may provide a new approach in monitoring disease progression in patients with RA. Adoption of these techniques may have implications for the management of patients with RA and for possible detection of the disease at an early stage.
1.1. Hand bone densitometry
Considerable advances have been made over the past two decades in developing radiological techniques for assessing bone density. However; all of these techniques have been utilized on aging-related osteoporosis, a pathological change involving general bone mineral reduction. Owing to the wide availability of DXA, recently published research describes the use of Bone Mineral Density (BMD) measurements in the hands of patients with chronic RA. Most published observations on RA have examined BMD changes, focusing on only the general bone loss around the joints. Quantification of the difference of bone loss between the juxta-articular bone and the shaft of tubular bones in the hands could be a sensitive index for quantitative analysis of RA patients. Hand BMD measurements offer an observer independent and reproducible means of quantifying the cumulative effects of local and systemic inflammation. The technique could be of use in the assessment of patients with early RA, in whom conventional measures of disease are not helpful until disease is (irreversibly) more advanced.
1.2. Hand radiographic absorptiometry
In conventional Radiographic Absorptiometry, radiographs of the hand are acquired with reference wedges placed on the films. The films are and subsequently analyzed using an optical densitometer. The resulting density values computed by the densitometer are calibrated relative to that of the reference wedge and are expressed in arbitrary units.
Recent improvements in hardware and software available for digital image processing have led to the quantitative assessment of radiological abnormalities in diagnostic radiology. Such improvements have also enabled introduction of several radiographic absorptiometry techniques. One such technique uses centralized analysis of hand radiographs and averages the BMD of the second to fourth middle phalanges. Another technique developed in Japan uses the diaphysis of the second metacarpal to determine BMD. A third technique developed in Europe measures the diaphysis and proximal metaphysics of the second middle phalanx. Based on published short-term precision errors, computer-assisted Radiographic Absorptiometry appears to be suitable for the measurements of the BMD of phalanges and metacarpals, and is used in several hundred centers worldwide.
In this work we present preliminary results of an ongoing research work aimed at developing an automated radiographic absorptiometry system for the assessment and monitoring of both BMD and soft tissue swelling in early stage RA. This paper focuses on the reproducibility and accuracy of the methodology being developed. The paper is organized as follows: the next section provides an overview of the image acquisition procedure. In section 3 the image analysis algorithms used in this work are presented. In section 4 we present results obtained by analyzing the data collected in a small reproducibility study involving 10 normal subjects.
2. IMAGE ACQUISITION
One key factor influencing the outcome of any radiographic absorptiometry technique is the standardization of the image acquisition technique. Variability in acquisition parameters can significantly affect the measured values. In order to carry out this work, a standard image acquisition protocol was defined. This protocol has been successfully used in earlier large scale multinational phase 3 clinical trials for Rheuatoid Arthrtis related drugs. Radiographs of the left and right hands are taken one
at a time. Templates were developed to guide the positioning of the hand with respect to the center of the x-ray beam. The X-ray beam was centered between the 2nd and 3rd metacarpo-phalangeal joints and angled at 90° to the film surface. This results in a tangential image of the joints. Improper beam centering generally results in overlapping joint margins. The X-ray exposure parameters were maintained constant for all subjects. All normal subjects were imaged at the same clinic at UCSF. In addition to providing a template for hand positioning, two sets of calibration wedges were also provided to the clinic. Each set of wedges consisted of one Acrylic wedge, for soft tissue and one Aluminum wedge for bone tissue. These wedges were custom designed for the purposes of this research work.
3.Enhancement Image and Restoration
The image at the left of Figure 1 has been corrupted by noise during the digitization process. The 'clean' image at the right of Figure 1 was obtained by applying a median filter to the image.
An image with poor contrast, such as the one at the left of Figure 2, can be improved by adjusting the image histogram to produce the image shown at the right of Figure 2.
3.IMAGEANALYSIS
One of the major difficulties in analyzing hand radiograph images is the high level of noise present in the images. Additionally the trabecular texture of the hand in the vicinity of the joints increases the noise in edge maps of this regions. Use of non-standard acquisition protocol can add additional challenges at it can result in further degradation of image quality. This last challenge is minimized in this work, as a standard image acquisition protocol is followed. Given a particular application varying degrees of accuracy in anatomy segmentation can be considered acceptable. For instance in detecting joint-space narrowing there is a need for accurate and reliable determination of the joint-space of any finger and the bone edges in this region. However, accurate delineation of the bone edges in the vicinity of the joint is not as relevant. Depending upon the application there can be additional constraints on performance issues as well. In an application for which off-line processing of the data is acceptable, more sophisticated algorithms can be employed. This particular application requires that the overall process be fast, accurate and reproducible enough for on-line processing. Accurate estimation of the bone edges in the middle shaft and in the joint vicinity of high relevance in this work. This is primarily because the disease progression follows different patterns in the joint area as compared to the middle phalange area. Also, the manifestations of the disease symptoms in its early stage have different effects on soft-tissue and bone as well, which require reliable segmentation of these two types of tissue at different time points. The algorithm for hand segmentation can be outlined into the following main stages:
• Hand outline delineation
• Joint identification
• Bone outline delineation
• Segmentation of soft tissue and bone
The first stage of the algorithm has been well studied in the literature and is not described here. The second stage can be more challenging, especially when dealing with hands of patients in advanced stage of disease progression. As this system will be applied to a patient population that is in their early disease stage it is expected that the joints will be well defined. The system provided to the radiologists allows them to adjust the location of the automatically identified joints. Results presented in this work were obtained by having the radiologists place control points to identify the joints, rather than having them automatically computed by the system.
3.1. Control point placement
A simple user interface was provided to enable placement of control points on the joints. This was primarily done to investigate the sensitivity of the system to the initial control point positioning, which in an automated system would invariably be the same for the same image. The user placed 16 control points on each image. These joints are show in Figure 4. In addition to placing the control points for the joints, the control points for the two wedges are also placed by the radiologists. For each wedge, six control points are placed with four at the corners and two in the middle. Once all the control points are placed, the remaining steps of the generalized algorithm stated above are carried out autonomously. The middle phalange or cortex control points are computed automatically and are located at the middle of straight line connecting the two joints, one above and one below the middle phalange. The diameters of the circular regions of interest placed around each joint are computed proportionally to the length
