A method for assessment of the shape of the proximal femur and its relationship to osteoporotic hip fracture

The shape of the proximal femur has been demonstrated to be important in the occurrence of fractures of the femoral neck. Unfortunately, multiple geometric measurements frequently used to describe this shape are highly correlated. A new method, active shape modeling (ASM) has been developed to quantify the morphology of the femur. This describes the shape in terms of orthogonal modes of variation that, consequently, are all independent. To test this method, digitized standard pelvic radiographs were obtained from 26 women who had suffered a hip fracture and compared with images from 24 age-matched controls with no fracture. All subjects also had their bone mineral density (BMD) measured at five sites using dual-energy X-ray absorptiometry. An ASM was developed and principal components analysis used to identify the modes which best described the shape. Discriminant analysis was used to determine which variable, or combination of variables, was best able to discriminate between the groups. ASM alone correctly identified 74% of the individuals and placed them in the appropriate group. Only one of the BMD values (Ward's triangle) achieved a higher value (82%). A combination of Ward's triangle BMD and ASM improved the accuracy to 90%. Geometric variables used in this study were weaker, correctly classifying less than 60% of the study group. Logistic regression showed that after adjustment for age, body mass index, and BMD, the ASM data was still independently associated with hip fracture (odds ratio (OR)=1.83, 95% confidence interval 1.08 to 3.11). The odds ratio was calculated relative to a 10% increase in the probability of belonging to the fracture group. Though these initial results were obtained from a limited data set, this study shows that ASM may be a powerful method to help identify individuals at risk of a hip fracture in the future.

Methods of measurement of image quality in teleultrasound

In recent years, teleultrasound has emerged as a potentially effective method of provision of ultrasound services to remote areas of the community. Transmission of teleultrasound, however, requires a high degree of compression when utilizing the relatively low bandwidth ISDN connections available within the UK. The actual amount of compression occurring varies both with the connection and the image content, often on a frame to frame basis. For this reason, traditional measurements of image quality using test tools are ineffective. Most studies to date have assessed equipment performance by testing diagnostic performance. Although this is the most important method of assessing the results of image degradation, few studies have attempted to perform physical measurements of degradation. The aims of this study were to measure the amount of degradation within a teleultrasound system and, using a number of methods, to ascertain the most useful indicators of loss of image quality. Groups of successive abdominal ultrasound images were captured and digitized both before and after degradation via an ISDN6 teleultrasound system. Normalized mean squared error (NMSE), pixel histograms, line and surface plots, Fourier spectra and contrast measurements were carried out on regions of interest in both sets of images. The NMSE was most sensitive to degradation, with alterations in the NMSE value depending on image content and degree of degradation. Fourier analysis was also indicative, showing consistent changes in high frequency content again varying with image content.

A comparison of digital and screen-film mammography using quality control phantoms

AIM

To compare the performance of a direct digital mammography system with normal-view and magnified-view conventional screen-film methods using quality control phantoms.

MATERIALS AND METHODS

Using a Siemens Mammomat((R))3000 and an Opdima((R))digital spot imaging and biopsy attachment, film and direct digital images of two phantoms [DuPont and TOR (MAM)] were obtained under normal operating conditions. These were assessed by three groups of observers with differing expertise - radiologists, radiographers and medical physicists. Each observer was asked to compare the direct digital image with films taken in standard view and magnified view, providing scores for object visibility and confidence. For the digital images, observers were allowed to vary the image presentation parameters.

RESULTS

Both phantoms showed that overall the direct digital view and the magnified view film performed significantly better (P < 0.05) than standard view film. For certain small or low contrast objects the differences became very highly significant (P < 0.001).

CONCLUSION

Only the TOR (MAM) phantom showed any significant difference between digital and magnified modalities, with magnified views performing better for fine, faint filaments and digital acquisition better for low contrast objects. Almost no difference existed between the three observer groups. Undrill, P. E. (2000). Clinical Radiology53, 782-790.

Analysis of trabecular bone structure using Fourier transforms and neural networks

Hip fracture due to osteoporosis (OP) and hip osteoarthritis (OA) are both important causes of locomotor morbidity in the elderly population. In osteoporosis, bone mass gradually decreases until the skeleton is too fragile to support the body and a fracture occurs, typically in the femur, wrist, or spine. In osteoarthritis, there is a proliferation of bone, leading to a stiffening of the tissue. Current clinical methods for assessment of bone changes in these disorders largely depend on assessing bone mineral density. However, this does not provide any information about bone structure which is considered to be an equally important factor in assessing bone quality. This paper presents a novel approach for computer analysis of trabecular (or cancellous) bone structure. The technique uses a Fourier transform to generate a "spectral fingerprint" of an image. Principal components analysis is then applied to identify key features from the Fourier transform and this information passed to a neural network for classification. Testing this on a series of 100 histological sections of trabecular bone from patients with OP and OA and a normal group correctly classified over 90% of the OP group with an overall accuracy of 77%-84%. Such high success rates on a small group suggest that this may provide a simple, but powerful, method for identifying alterations in bone structure.

Split-and-merge segmentation of magnetic resonance medical images: performance evaluation and extension to three dimensions

Intensity- or edge-based methods of segmentation are often insufficiently robust to be applied to images containing complex anatomical objects, such as those seen in high-resolution magnetic resonance imaging systems. Split-and-merge techniques attempt to overcome these difficulties by using homogeneity measures. Simple modifications to the basic 2D split-and-merge method, based on the principles of simulated annealing and controlled boundary elimination, are developed and discussed. Simulated annealing reduced the number of regions by 22% with a further reduction of 21% achieved through boundary elimination. Smoother regional boundaries are also produced. These methods are extended to true 3D and quantitatively compared with their 2D counterparts. The main advantage of 3D methods is that they produce segmented volumes by directly preserving the connectivity between slices, whereas in 2D, segments have to be grouped together in a post-split-and-merge process. Finally, the properties of the 3D approach are demonstrated by the automatic quantitation of brain ventricle volume, producing estimates to within 7% of validated manual methods.

A comparison of computer based classification methods applied to the detection of microaneurysms in ophthalmic fluorescein angiograms

We compared the performance of three computer based classification methods when applied to the problem of detecting microaneurysms on digitised angiographic images of the retina. An automated image processing system segmented 'candidate' objects (microaneurysms or spurious objects), and produced a list of features on each candidate for use by the classifiers. We compared an empirically derived rule based system with two automated methods, linear discriminant analysis and a learning vector quantiser artificial neural network, to classify the objects as microaneurysms or otherwise. ROC analysis shows that the rule based system gave a higher performance than the other methods (p = 0.92) although a much greater development time is required.

The use of texture analysis to delineate suspicious masses in mammography

In mammography, national breast screening programmes have lead to a large increase in the number of mammograms needing to be studied by radiologists. Lesion indicators can be pointlike as in microcalcifications or extended as in stellate (spiculate) lesions or regular masses. Texture analysis has been proposed as a promising method for studying radiographic images in relation to the quantitation of extended objects. Filters have been designed, which may be used to segment or classify an image using textural features, and these have been reported as being of value in automatic mammographic glandular tissue classification. The work reported here suggests the incorporation of additional steps of image processing in an attempt to improve the performance of these filters in the quantitation of lesions. By deriving approximate outlines, which are used to identify suspicious regions, the investigation illustrates the properties of one of the filters. After applying the method to a small prediagnosed database of stellate lesions and regular masses, the results show that the filter is able to outline the malignant masses in all cases presented. The erroneous areas extracted are small for the initial part of the work, which deals with 256 x 256 pixel image extracts, though slightly larger in some cases when the whole mammogram is considered. Simple methods for the removal of these artefacts are proposed. For each non-suspicious case studied, the sum of any false positive areas is statistically insignificant when compared with the regions correctly outlined in the prediagnosed instances.

The use of texture analysis to delineate suspicious masses in mammography

In mammography, national breast screening programmes have lead to a large increase in the number of mammograms needing to be studied by radiologists. Lesion indicators can be pointlike as in microcalcifications or extended as in stellate (spiculate) lesions or regular masses. Texture analysis has been proposed as a promising method for studying radiographic images in relation to the quantitation of extended objects. Filters have been designed, which may be used to segment or classify an image using textural features, and these have been reported as being of value in automatic mammographic glandular tissue classification. The work reported here suggests the incorporation of additional steps of image processing in an attempt to improve the performance of these filters in the quantitation of lesions. By deriving approximate outlines, which are used to identify suspicious regions, the investigation illustrates the properties of one of the filters. After applying the method to a small prediagnosed database of stellate lesions and regular masses, the results show that the filter is able to outline the malignant masses in all cases presented. The erroneous areas extracted are small for the initial part of the work, which deals with 256 x 256 pixel image extracts, though slightly larger in some cases when the whole mammogram is considered. Simple methods for the removal of these artefacts are proposed. For each non-suspicious case studied, the sum of any false positive areas is statistically insignificant when compared with the regions correctly outlined in the prediagnosed instances.

A transputer based three-dimensional dynamic cardiac imaging system

In Aberdeen, a single-section transverse emission scanner has been adapted to trigger off patient ECG signals, allowing for the acquisition of gated blood pool tomograms at a number of slice levels through the patient's heart. This paper describes a system for the routine generation and display of surface rendered images derived from this data using a transputer based hardware system. Surface rendering algorithms have been implemented to provide an indication of the distribution of the cardiac blood pool in three dimensions, whilst the additional use of colour and/or cine sequences provide a succinct method of representing the extra information provided by gated acquisition. The transputer provides sufficient computing power to produce rendered views at a rate of about 1 frame per second, thus putting view selection fully under operator control. The success of the system described is reflected in its routine use in a busy clinical department.

The Aberdeen Mark II single-photon-emission tomographic scanner: specification and some clinical applications

The construction, operation and physical characteristics of a single-section multi-detector single-photon-emission scanner are described. The machine has 24 detectors arranged along the sides of a square. Movements and data collection are under the control of a series of distributed microprocessors. Both head and trunk tomograms can be produced. The spatial resolution at the collimator focus is 9 mm in the transverse plane, and the effective slice thickness is 14 mm. The volume sensitivity is 300 counts/s kBq ml with a 20 cm diameter cylindrical phantom filled with 99Tcm solution. The application of this machine to the examination of the brain, liver and heart has been found to be clinically useful.

A simple filtering routine for radionuclide bone images

The evaluation is carried out of a simple filtering routine for improving the detectability of low contrast lesions in bone scans. With this routine, the unprocessed image is displayed together with three filtered versions, the final decision on the presence of abnormalities being made by comparing the four images. Emphasis is placed on producing a routine which is quick in implementation and, therefore, only small filter arrays are used. Using a routine consisting of a low-pass, a median, and a differential filter, 73% more lesions are detected when the filtered images are also used. The whole routine takes less than 1 min to run.

A clinical trial of the value of a tomographic section view to identify liver abnormalities by radionuclide imaging, with special reference to metastatic disease

A prospective trial was carried out to assess the value of a radionuclide transverse section view in addition to conventional radionuclide scans of the liver. Reports for 373 patients were analysed and compared with the final independent clinical diagnosis. Although several individual cases were more accurately diagnosed using the tomographic section view, the trial did not demonstrate a significant increase in accuracy of diagnosis in liver disease. Some possible reasons for this are discussed, mainly the difficulty of identifying and quantifying truly abnormal features of liver radioisotope scans.

Cranial scintigraphy: value of adding emission computed tomographic sections to conventional pertechnetate images (512 cases)

A prospective trial was carried out to assess the contribution made by a radionuclide transverse-section view to conventional radionuclide scans of the brain. Each set of scans was reported by two independent teams of observers, but only one team viewed the tomography section. An abnormality rating was used to decide whether a set of scans was positive or normal. The reports for 512 patients were analyzed and compared with the final independent diagnosis. Greater accuracy of diagnosis with the section view was significant at the 0.1% level; the results showed that failure to detect tumors was almost halved while there was a 16% improvement for the detection of infarcts. Abnormality ratings were also used to plot ROC curves and rating curves; these showed that the addition of the tomographic view increased markedly the proportion of true positives without any increase in the proportion of false-positive reports.

Incorporation of a SMAC analyzer into the data-processing procedures of a computer-assisted laboratory

This paper describes the incorporation of a SMAC (Technicon) analyzer into data-processing techniques that have been developed on existing computer hardware during several years. The SMAC system is interfaced directly to a small computer, and suitable peripherals produce a manageable form of result tabulation for subsequent reporting, as well as provide quality-control information to the SMAC operators in real time. The design is such as to facilitate the performance analysis of the SMAC system during its initiation period and during normal service operation.

Incorporation of a SMAC analyzer into the data-processing procedures of a computer-assisted laboratory

This paper describes the incorporation of a SMAC (Technicon) analyzer into data-processing techniques that have been developed on existing computer hardware during several years. The SMAC system is interfaced directly to a small computer, and suitable peripherals produce a manageable form of result tabulation for subsequent reporting, as well as provide quality-control information to the SMAC operators in real time. The design is such as to facilitate the performance analysis of the SMAC system during its initiation period and during normal service operation.

An approach to quality and performance control in a computer-assisted clinical chemistry laboratory

A locally developed, computer-based clinical chemistry laboratory system has been in operation since 1970. This utilises a Digital Equipment Co Ltd PDP 12 and an interconnected PDP 8/F computer. Details are presented of the performance and quality control techniques incorporated into the system. Laboratory performance is assessed through analysis of results from fixed-level control sera as well as from cumulative sum methods. At a simple level the presentation may be considered purely indicative, while at a more sophisticated level statistical concepts have been introduced to aid the laboratory controller in decision-making processes.

An on-line work scheduling and ward requesting pattern information system

The organisational efficiency of sample admission to the routine laboratory and the allocation to test methods has been improved by a scheduling system based on an on-line mini-computer. A flexible system has been constructed using multiple visual displays, which produce work allocation lists. Computer techniques have been developed which enable maximal use of the on-line system while retaining the general flexibility of high-level language coding, as implemented on an unsophisticated 12-bit processor. The file structure has been designed to provide cumulative information on the requesting patterns appropriate to any of the many sources of samples to the laboratory, and to allow the automatic merging of worksheet and laboratory data, the latter being generated and transmitted from a second mini-computer system.