Development of phantoms for spiral CT.

PURPOSE: This paper reports on the development of a new phantom for spiral CT. The phantom meets the increased demands on phantom z-axis uniformity in order that objects from the CT slice, immediately above and below the CT slice of interest, do not contribute perturbing information to the reconstructed CT slice. MATERIAL AND METHODS: The phantom depends on formulation of tissue-like materials that can be cast and produced in both geometric and anthropomorphic shapes with sufficient z-axis length to enable unperturbed CT slices of test objects of interest. These materials are then used to produce a series of test objects of CT image quality including low contrast samples that do not require volume averaging or mixing of solutions, and that can reflect sub-slice thickness test objects and supra-slice thickness test objects. RESULTS: The overall phantom and its individual test objects provides meaningful tests of spiral CT image quality including slice sensitivity, CT number linearity and tests of high and low contrast resolution. Schematic designs and actual CT scans are shown. CONCLUSION: The new spiral phantom appears to meet the increased demands of spiral CT on phantom design, particularly z-axis length, and requirements for low contrast resolution test objects.

Objective analysis of ultrasound images by use of a computational observer.

The design, implementation, and testing of a computational observer method for objective evaluation of ultrasound images are presented. The method uses digitized ultrasound B-scan images of a test phantom (the contrast-detail phantom), and is able to calculate the detectability of a target (signal) from its background (background noise). A quantitative detectability index, based on the measured signal-to-noise ratio of the image data, that is measured for both the human and the computational observer on the same scale is generated. It is shown that the computational observer (CO) method may be a more useful, objective way of evaluating ulrasound images and imaging systems than methods that rely solely on human observers. It may also be applicable to other types (i.e. other than ultrasound) of imaging systems which produce noisy images. The relevance of the CO method when compared to human observer two-alternative-forced-choice (2AFC) readings of the same data by showing a high correlation between the CO detectability results and those of human observers for the same set of images. The method is (1) quantitative, (2) reproducible, (3) absolute, (4) takes into account, and can calculate the value of TPF and FPF for each target, for the given system, and (5) speeds up the evaluation of an image or imaging system (compared to using human observers), given the right conditions and equipment.

Precise measurement of vertebral bone density using computed tomography without the use of an external reference phantom.

Bone density measurement by quantitative computed tomography (QCT) commonly uses an external reference phantom to decrease scan-to-scan and scanner-to-scanner variability. However, the peripheral location of these phantoms and other phantom variables is also responsible for a measurable degradation in accuracy and precision. Due to non-uniform artifacts such as beam hardening, scatter, and volume averaging, the ideal reference phantom should be as close to the target tissue as possible. This investigation developed and tested a computer program that uses paraspinal muscle and fat tissue as internal reference standards in an effort to eliminate the need for an external phantom. Because of their proximity, these internal reference tissues can be assumed to reflect more accurately the local changes in the x-ray spectra and scatter distribution at the target tissue. A user interactive computerized histogram plotting technique enabled the derivation of reproducible CT numbers for muscle, fat, and trabecular bone. Preliminary results indicate that the use of internal reference tissues with the histogram technique may improve reproducibility of scan-to-scan measurements as well as inter-scanner precision. Reproducibility studies on 165 images with intentional region-of-interest (ROI) mispositioning of 1.5, 2.5, or 3.5 mm yielded a precision of better than 1% for normals and 1% to 2% for osteoporotic patients--a twofold improvement over the precision from similar tests using the standard technique with an external reference phantom. Such improvements in precision are essential for QCT to be clinically useful as a noninvasive modality for measurement of the very small annual changes in bone mineral density.

Phantoms for specifications and quality assurance of MR imaging scanners.

The accompanying paper examines the subject of NMR phantoms. The paper reports on initial experience with existing phantoms and reviews proposals from various standards groups and professional organizations. Many image tests are illustrated by existing vendor phantoms. The paper concludes that many phantoms already meet or exceed most of the suggestions for tests of classical imaging parameters, which can be pursued by first order adaptations of CT phantoms. The paper does, however, point out the limitations of existing phantoms and raise the possibility of developing phantoms that more accurately reflect human shapes and cavities, and which present more realistic resistive losses, and T1 and T2 values.

A new software correction approach to volume averaging artifacts in CT.

The nature of artifacts in computed tomography, due to nonlinear partial volume effects, is briefly reviewed. A methodology for correction of these artifacts proposed. This methodology utilizes a correction algorithm for post-processing the original reconstructed CT image. The algorithm utilizes local CT values to predict the probability of volume averaging of bone, air and tissue, and incorporates information from the original image on the spatial extent of the averaging, to correct the nonlinear effects. The algorithm is successfully demonstrated on mathematical phantoms, wherein volume averaging can be introduced over central targets and/or peripheral annuli of tissue, bone, and air. The algorithm is also demonstrated on actual CT brain scans but only for its qualitative effects. It is pointed out that higher order corrections, to be reported in future publications, will further correct the actual clinical scans.

Factors related to low contrast resolution in CT scanners.

The concept of "low contrast resolution" in CT scanners is examined. Several questions are raised, not only about limitations to interpretation of figures of merit of low contrast resolution, such as contrast-detail diagrams; but also, interpretation of related CT physical performance parameters, such as: noise, dose, spatial resolution and slice width (or sensitivity profile). Caveats are raised concerning interpretation of contrast-detail studies and the degree to which first-order observer studies are reflective of the actual clinical situation. Particular care is suggested in interpretating contrast-detail diagrams in which sensitivity, specificity and the number of target signals may be varying.

Comparative image aspects of radiography and computed tomography.

Fundamental image descriptors, including contrast, noise, and resolution, were appropriate for computed tomography. These variables were then used to predict limiting system performance as applied to a model for detection of simple tumor-like signals in a homogeneous surround. The results predict that any intrinsic physical advantages of computed tomography may be restricted to a fairly narrow range of tumor diameters, from a few millimeters to a few centimeters, depending on the relative dose levels and scatter levels present in each imaging system. The results indicate high probable yield for radiographic image manipulation involving independent or, preferably, mutual manipulation of scatter levels, contrast enhancement, kVp techniques, and subtraction techniques, where applied to reduce the highly structured backgrounds that can increase the complexity of diagnosis.

Theoretical limitations in the immunodiagnostic imaging of cancer with computed tomography and nuclear scanning.

In order to help assess the feasibility of using immunologically tagged agents to render tumors detectable with current computed tomographic and nuclear scanners, mathematical formulations were developed to determine the theoretical limits of tumor detection relative to size and depth of the lesions. The results of our analysis suggest that visualization with computed tomography of a tiny tumor (1 sq mm, cross-sectional area) would require binding in the order of 2 x 10(5) iodine atoms/antigenic site, while imaging of a very large (900-sq mm) tumor would require approximately 10(4) atoms/site. Very low energy scanners might reduce these discouraging estimates by an order of 10(2). The immunological imaging of tumors with nuclear scanning appears quite feasible from our formulations, as has been demonstrated by others clinically. Small (1-sq cm) and deep (greater than or equal to 5-cm) tumors appear detectable with uptake ratios of the order of 5 or higher, which seem to be attainable currently. Smaller and deeper tumors require much higher uptake ratios to be detected.

Familial and acquired paroxysmal dyskinesias. A proposed classification with delineation of clinical features.

On a clinical basis the paroxysmal dyskinesias can be classified into two distinct categories--familial and acquired. The former begins in childhood and the dyskinesia may or may not be induced by sudden movements (kinesigenic or nonkinesigenic forms). In the familial kinesigenic form, the movements are brief, usually occur daily, and respond readily to anticonvulsants. This form has an autosomal dominant or recessive mode of inheritance. In the familial nonkinesigenic form, the movements are of longer duration, occur less frequently, and rarely respond to anticonvulsants. This form has a clear autosomal dominant mode of inheritance. The etiology is obscure. The acquired form of paroxysmal dyskinesia has a later onset and is an expression of an underlying neurological or metabolic disease. Some cases of acquired paroxysmal dyskinesia are manifestations of unusual forms of epilepsy. In these cases the differential diagnosis may be extremely difficult and must be based on EEG findings during an ictal episode.

Radiographic imaging using an effective rotationally symmetrical focal spot.

A mathematical model is presented for predicting the effect of rotation on certain types os asymmetric focal spot intensity distributions. Experimental rotational techniques are used to obtain rotationally symmetric, gaussian-like focal spot distributions from existing asymmetric distributions. Radiographic examples are shown which illustrate the resulting imaging effect of such rotationally symmetric focal spots compared to existing focal spots. In particular, images of resolution test patterns and simulated blood vessels are used to examine some of the radiographic consequences (contrast transfer and spurious imaging) of the use of differing types of focal spot intensity distributions, and to question the utility of single number characterizations of asymmetric focal spots.

Potential artifacts associated with the scanning pattern of the EMI scanner.

Physical and geometrical considerations of the EMI scanner are used to predict that regions of the head can be missed by adopting a scanner procedure based on the nominal EMI slice thickness. Using a nominal slice thickness of 1.3 cm and effective focal spot length of 12 mm, one can predict that in the central portion of the head, up to 8% of the region scanned may not be imaged. Moreover, it is shown that simultaneous monitoring of a single focal spot by two adjacent detectors, as in the EMI system, would lead to significant overlap between the two adjacent EMI slices resulting in dual imaging. Experimental examples are shown to illustrate these two artifacts in the EMI scanner.

Visual detection and localization of radiographic images.

In conventional receiver-operating-characteristic (ROC) curve analysis of visual detection performance, the observer is credited with a true-positive response if a visual signal is present somewhere in a radiograph called "positive" by the observer; however, the measured true-positive rate can be different for a given false-positive rate if the observer is required to identify the correct location of the visual signal in order to receive credit for a true-positive response. The authors describe and have confirmed experimentally a model which can be used to predict observer performance in an experiment requiring both detection and localization on the basis of the conventional ROC curve determined in a detection experiment. Implications for the use of signal detection theory in the assessment of radiographic image quality are discussed.

The use of ROC curves in testing the proficiency of individuals in classifying pneumoconiosis.

During a test of the ability to classify the profusion characteristics of pneumoconiosis recorded in chest radiographs, a constant decision criterion level needs to be maintained or false conclusions regarding the observer's receiver operating characteristic (ROC) curve may result.