Medical Imaging : CT

Medical Imaging Portfolios


Image Analysis Method Normalizes Skeleton; Eases Physician Burden

UW–Madison researchers have developed a statistically optimized regional thresholding (SORT) method, which establishes the first set of optimized bone-by-bone thresholds to detect lesions throughout the entire skeleton in NaF PET/CT images.

Their method is based on differentiating diseased from healthy signals in different skeletal regions. They developed a standardized skeleton ‘template’ that reduces image features related to normal physiology and accentuates features related to disease. To achieve this, they analyzed multiple healthy individuals with respect to radioactive tracer uptake and established anatomy-dependent background signal thresholds. These values can serve to statistically select the best thresholds for identifying lesions in different skeletal regions.

After the determination of thresholds, a normalized image is produced that can be more easily analyzed by the physician, having had standard variations removed so that only disease-based differences are evident. The improved image dataset may also be used for better automatic analysis of lesion size, location and change.

Method for Data-Consistency Preparation and Improved Image Reconstruction

Recognizing the link between data consistency and artifact mitigation, UW–Madison professor Guang-Hong Chen has developed an improved processing method that is applicable across modalities, including CT imaging, PET, SPECT and MRI.

The new method provides a practical means to define a data inconsistency metric (DIM) that can be used to locally characterize the inconsistency level of each acquired datum or a view of acquired data. The DIM can be used in a data classification technique to select an optimal data set with a minimal data inconsistency level to reconstruct images with minimal artifact contamination.

The acquired datasets can be classified into one or more subsets based upon the value of DIM. After data classification, a reconstruction technique, such as the SMART-RECON algorithm pioneered by Prof. Chen, can be used to reconstruct these sub-images. Each sub-image is consistent with the subset of the projection view angles for a given range of DIM values. The result is substantially improved images.

Eliminating CT Image Artifacts Using Artifact ‘Dictionaries’

A UW–Madison researcher and industry collaborator have developed a new method that would be incorporated in software and used prior to a radiologist reviewing the CT images. The method involves two key steps. In the first step, the image is divided into many small patches, which may or may not overlap. The size and/or location of the patches may be anatomy- and/or location-dependent.

In the next step, each small patch is decomposed into two sub-patches with one sub-patch corresponding to one or more artifacts and the other devoid of artifacts, or having reduced artifacts. After each small patch is separated into artifacts and non-artifacts, the artifact patches and non-artifact patches may be recombined to generate an artifact image and an artifact-free image (or reduced artifact image).

To achieve this, two dictionaries are constructed: an artifact dictionary and a non-artifact dictionary. Using such constructed dual dictionaries a priori, each small image patch may be decomposed into an artifact patch and a non-artifact patch. It may be noted that multiple passes or iterations may be made on a given patch or patches to remove more than one type of artifact.

Faster, Higher Quality Medical Imaging

UW–Madison researchers have developed a reconstruction technique that uses a non-patient-specific signal model (e.g., a physical or physiological model) to improve image quality without compromising accuracy.

While other methods make use of such analytical models in the post-processing stage, the new technique utilizes the model earlier in the process, yielding clinically useful images from highly undersampled data. The reconstruction process is designed to accommodate deviations from the model when appropriate.

Improved CT Imaging with Multisource X-Ray Tube

UW–Madison researchers have developed a compact, multisource X-ray tube for use in CT imaging. The new tubes can deliver high current from an arbitrary number of focal spots. Utilizing a modular design, the tubes may be arranged in a variety of configurations to suit a particular application.

A module consists of a series of electron emitters that can be switched on and off at high frequency, and are directed towards a single stationary target that is actively cooled. A large voltage between the emitters and the target accelerates the electrons to high energy. Upon impact with the target, the electrons produce an X-ray spectrum. An electromagnet is used to ‘sweep’ the multiple electron beams over the cooled target.

In effect, X-rays are generated around a patient in rapid succession, much faster than the mechanical motion of a rotating gantry.

Multi-Angle CT Scanning of Stationary Patients

Researchers from UW–Madison and the Morgridge Institute for Research have developed a gantry support structure that uses articulating arms to scan humans or animals sitting, standing or lying down.

Their articulating robotic arm system (called ARMS) controls the motion of a standard high-speed CT scanner. ARMS moves around a patient, producing complex trajectories while allowing the patient to remain stationary.

The articulating arms have motorized shoulder, wrist and elbow joints to support greater motion capabilities. The system is suitable for scanning humans as well as large animals, and may be mounted on the floor, ceiling or wall for added versatility.

Dual-Energy CT Cuts Costs, Radiation Dose

A UW–Madison researcher has developed a system for generating multi-energy CT images using a single-energy, polychromatic X-ray source. ‘Polychromatic’ means the X-rays have more than one wavelength.

In the new process, the polychromatic radiation is delivered to a patient and received by a detector array, which generates attenuation data. This data is segmented based on several component criteria. Ultimately, various forms of data are synthesized and used to reconstruct a number of multi-energy images, including separable images weighted for each of the component criteria.

X-Ray Phase Contrast Imaging Using Standard Equipment

A UW–Madison researcher has developed a method for generating X-ray phase contrast images from conventional X-ray attenuation data.

First, calibration factors are obtained using a phantom. The patient or object then is X-rayed to acquire attenuation data at two different energy levels. Images are reconstructed at the different energy levels to produce spatial maps. Based on the calibration factors and spatial maps, a phase contrast image can be produced.

Statistical Imaging Reconstruction Is Faster, Cuts Noise

A UW–Madison researchers has developed an iterative reconstruction method that simultaneously achieves high convergence speed and high parallelizability. The method can work with various medical imaging systems, including CT, magnetic resonance imaging (MRI), X-ray angiography and positron emission tomography (PET).

In general, a nonlinear reconstruction problem is decomposed into separate linear sub-problems that can be solved more efficiently. The statistical image reconstruction process is decomposed into a statistically weighted algebraic reconstruction update sequence. After this step, the image is de-noised using a regularization function.

Evaluating Systemic Cancers

UW–Madison researchers have developed a technique for extending molecular and functional imaging (e.g., PET, fMRI) assessment of the total disease and disease heterogeneity to a variety of different cancers, including systemic types throughout the body.

The method uses a combination of anatomical and functional masking to isolate multiple dispersed lesions from surrounding tissue. In this way, automatic identification tools can target likely tissue on a case-by-case basis, as guided by information about the type of cancer and imaging materials.

First, a patient is administered an imaging agent that identifies tumor tissue. After scanning, a program helps identify and measure the progression of multiple tumor locations based on how and where the agent is taken up. A color-coded output shows measurements at different locations.

Simultaneous Image Reconstruction and Artifact Reduction

A UW–Madison researcher has developed a system for reconstructing images with different levels of artifacts. In this way, a ‘target image’ with the lowest level of problems will be produced simultaneously with an ‘artifact image’ that depicts primarily artifacts.

The method works by automatically and iteratively producing multiple images from one set of data, with the multiple images corresponding to different data consistency levels.

Once a subject is scanned, an image matrix is initialized having columns that correspond to different images. At least one image then is reconstructed by minimizing a matrix rank. The ranking is constrained according to a consistency condition that promotes the forward projection of each column to be consistent with a different subset of the acquired data.

Statistical Noise Map for Reducing X-Ray Exposure

UW–Madison researchers have developed a system and method for estimating a statistical noise map from a single X-ray exposure. This map accounts for noise acquired with X-ray imaging systems, including computed tomography (CT), tomosynthesis and C-arm systems.

The method reconstructs an image from acquired data using any standard filtered back projection (FBP) algorithm. This image is used as a baseline to estimate a noise standard distribution map. The raw projection data represents a typical measurement among many repeated measurements under the same experimental conditions. Therefore, this data can be used to generate several (e.g., 20 or more) noisy data sets.

These data sets are used to reconstruct noisy images that can be subtracted from the original image, resulting in a statistical noise map. This map accounts for a physical model of noise.

Medical Imaging with Better Temporal Fidelity Can Streamline Stroke Care

UW–Madison researchers have developed a method that increases temporal fidelity, sampling density and/or reduces noise of image frames obtained with a system such as CT, MRI or X-ray c-arm. After the images are acquired, a window function is selected and temporally deconvolves the image frames using a minimization technique. A temporal sampling density also may be selected and used in the temporal deconvolution.

Faster, Better Quality Medical Imaging by Constrained Reconstruction

UW–Madison researchers have developed a modified algorithm for medical image reconstruction that increases reconstruction speed, improves image quality and provides more accurate results. The algorithm constrains images to be consistent with a signal model, which relates image intensity values to free and control parameters such as relaxation time and multiple echo or inversion times, respectively.

The signal model may be analytical or approximate—learned from acquired image data, as is done in the case of time-resolved MRI. The model consistency condition may be enforced using an operator that projects an image estimate onto the space of all functions satisfying the signal model.

Reducing Image Noise and X-ray Dose in Spectral CT

A UW–Madison researcher and others have developed a method to improve spectral CT by allowing users to select an optimal energy bin configuration. Specifically, the width, location and number of bins can be chosen so that the best material information can be gathered without the images succumbing to noise.

In the method, the multi-energy CT system acquires a series of energy data sets. Each data set is associated with at least one energy bin. A ‘conglomerate image’ is created, which uses most or all of the X-ray photon data spanning the energy bins. Then, an energy series of images can be produced, with each image in the series corresponding to at least one of the data sets. In this way, noise is reduced in each energy-specific image, not just in the final processed image.

The method may employ a previously developed reconstruction algorithm known as HYPR (HighlY constrained back-PRojection).

“DR-PICCS” – Dose Reduction Using PICCS Image Reconstruction Algorithms

UW-Madison researchers have developed a method using existing “PICCS” (prior image constrained compressed sensing) image reconstruction algorithms to reduce radiation dose while attaining quality images with high SNR. Multiple slices of an image volume are collected and then averaged together to create a single thick slice, known as the “prior image,” with high SNR but lacking detailed anatomical structures. The PICCS algorithm then is used to reconstruct each image slice with the original slice thickness using the prior image. The resulting final image has the equivalent image noise variance level of the prior image, the high spatial resolution of the acquired image is preserved and the anatomical features will be detailed.

Noise variance in the final reconstructed image is improved by a factor of approximately the number of slices included in the prior image. Using this method, the patient receives a reduced dose of radiation while the radiologist acquires final images of the same quality currently attained with higher radiation levels.

Hybrid Method for Prior Image Reconstruction in Cardiac Cone-Beam Computed Tomography

UW-Madison researchers have developed a method for producing a prior image from truncated cone beam projection data such that quality images can be reconstructed using methods such as PICCS. This image reconstruction method is applicable to cardiac cone beam X-ray computed tomography.

The image reconstruction method uses a set of truncated cone beam image data obtained using cardiac flat panel detectors to reconstruct a high quality image in two steps. In the first step, an iterative algorithm is used to reconstruct a prior image from the truncated data without the use of ECG gating, a technique used to trigger image acquisition with a patient’s ECG signal. In the second step, the reconstructed prior image is used in a PICCS algorithm to reconstruct each individual cardiac phase. The objective function of the PICCS algorithm is modified to incorporate the conditions used in the first step to overcome data truncation. Together, the method comprises a hybrid PICCS algorithm that enables computed tomography with cardiac cone-beam X-ray systems, resulting in satisfactory reconstructed imagery.

Range Adjusted Dynamic Image Construction Algorithm (RADICAL) Method to Suppress Artifacts in X-Ray Imaging

A UW–Madison researcher has developed a method for substantially suppressing streak artifacts in images produced with an X-ray imaging system without significant loss of information. The Range Adjusted Dynamic Image Construction Algorithm (RADICAL) approach reconstructs an image by weighting the coefficient values of highly attenuating regions so that values above a user specified threshold contribute less to a given location in the image than those below the threshold. This method removes much of the metal artifacts on CT imaging, resulting in a more accurate reconstructed image. In addition, it may limit other inconsistent data artifacts.

Cardiac Image Reconstruction with Improved Temporal Resolution

UW–Madison researchers now have developed a ‘prior image’ method for reconstructing dynamic and undersampled data. The method is applicable to a number of different modalities including CT, X-ray C-arm imaging, MDCT, magnetic resonance imaging (MRI), positron emission tomography (PET) and single photon emission CT (SPECT).

Specified for each system, in general the method combines image data from current and prior time frames, like heartbeat phases. A limited amount of additional image data is incorporated into the consistency condition imposed during prior-image constrained-image reconstruction.

Real-Time Progressive Medical Image Reconstruction Method for Time-Resolved Data

UW–Madison researchers have developed a method for medical image reconstruction that delivers quality images from time-resolved image data in real time. The method provides accurate images with increased signal-to-noise ratio and temporal resolution while minimizing a patient’s exposure to X-ray radiation.

The method starts with two related images that can be neighboring images from a time series image set or images from the same phase point during repeated motion, such as breathing or the beating of the heart. The images are subtracted to get an image of the difference between the two, which undergoes a sparsifying transformation to reconstruct the final image of interest. All of this is done in real time to achieve more dynamic medical imaging, such as image guided interventional procedures.

Image Reconstruction Method for High Temporal Resolution Image Guided Radiaton Therapy

UW–Madison researchers have developed a medical image reconstruction method designed to increase temporal resolution, while increasing accuracy and reducing the radiation dose to the patient. The method may be applicable to numerous imaging techniques including magnetic resonance imaging (MRI), X-ray computed tomography (CT), positron emission tomography (PET) and single photon emission computed tomography (SPECT).

Acquired data is used to reconstruct a “sparsifying image.” A “correction image” is iteratively determined and subtracted from the sparsifying image to produce a quality image.

The technique also can be applied to current IGRT techniques to increase the accuracy of radiation delivery. Sparsifying images are obtained for a specific phase of the respiratory cycle to more accurately determine the motion characteristics of the target tumor and increase the temporal resolution.

High Temporal Resolution Cardiac CT Imaging with Slowed Gantry Speed

UW-Madison researchers have developed an application of the PICCS image reconstruction method (see WARF reference number P08127US) for producing a time series of images with a higher temporal resolution than the temporal resolution at which the image data was acquired. In cardiac imaging, this allows use of slow gantry rotation for improved image resolution instead of continuing to increase the speed of gantry rotation, which is mechanically challenging.

For high temporal resolution cardiac imaging, a “cone-beam” arrangement such that the focal spot of the X-ray source and the detector define a cone-shaped beam of X-rays is used. The gantry rotation time is counter-intuitively slowed to about 10 seconds. This rotation time enables a single breath-hold for most cardiac patients, which reduces motion during imaging. During the cone-beam CT data acquisition, the ECG-signal will be recorded as 400 to 600 views of the cone-beam projection are simultaneously acquired during each gantry rotation. The acquired data will be used to reconstruct a “prior” image containing the heart that does not contain dynamic information and possibly contains motion-induced streaks. Next, the acquired projection data are “gated” using the ECG data so that there is one projection per heart beat and images can be reconstructed using data from each gated “window”. This allows the PICCS algorithm to accurately reconstruct each cardiac phase. The resulting images have an ultra-high temporal resolution about 20 times better than images obtained using state-of-the-art CT scanners with increased gantry speeds.

Prior Image Constrained Compressed Sensing (PICCS)

UW-Madison researchers have developed a method for reconstructing a high quality image from undersampled image data that is applicable to a number of imaging modalities including CT, MRI and positron emission tomography (PET).

A seed image is acquired by using a prescan, reconstructing a high signal-to-noise (SNR) ratio image from data acquired through any modality, or from a fully sampled data set. This prior seed image is used to iteratively reconstruct a final output image from an undersampled data set taken of the same anatomical structure as the seed image. The high SNR seed image guides a mathematical manipulation of the data set, resulting in a high quality image constrained by the original high SNR image. The method typically requires only two to five iterations to achieve clinically useful images, resulting in a convergence speed much faster than any known iterative image reconstruction methods.

Method and Apparatus for Low Dose Computed Tomography

UW-Madison researchers have developed a CT machine capable of minimizing the radiation dose while providing an image of desirable quality. The contribution of each radiation beamlet to the quality of the resulting tomographic image varies as a complex function of the internal structure of the patient. Although the structure of the patient is generally unknown, in many cases there is sufficient a priori knowledge about the patient to intelligently select beamlets based on how important each beamlet is to the quality of the image. The CT machine of this invention features a controller that uses a stored model of the patient to control the intensity of radiation in each beamlet based on a calculated contribution of the beamlet to image quality.

Ultra Low Radiation Dose Computed Tomography Scanner for X-ray Mammography

A UW–Madison researcher has developed a line scan X-ray cone beam CT scanner system and method with an ultra low radiation dose to provide high in-plane spatial resolution and excellent low contrast resolution. The system is made up of an x-ray cone beam source and a 2-D X-ray detector array. Both move linearly on opposing sides of the breast in opposite directions to capture data from various angles. They do not move in a circular pattern around the breast.

An image is reconstructed with the data from a limited amount of angles, which results in a significantly lower X-ray dose to the patient. The cone beam data is converted into a parallel beam projection data set to create an image using a novel image reconstruction method. This method determines the image quality by using a determined metric to measure it against with the image quality continually increased until a preset quality is met. The limited number of angles required for this method leads to an approximate 20- fold reduction in radiation dose.

Image Reconstruction System and Method to Reduce Artifacts from 3-D Cone-Beam CT Systems

A UW–Madison researcher has developed a system and method for producing 3-D medical images from cone-beam CT scans that are significantly free of image artifacts due to missing data. The system includes a computer readable storage medium that carries out the method of improving the image.

The system takes the incomplete data acquired from a CT scan and creates an image of the volume of interest with artifacts attributed to the missing data. This image is optimized a desired amount and then reprojected onto a virtual scan path that acquires virtual data to take the place of the missing data. The originally acquired data and virtual data are combined to produce an improved image with a significantly reduced amount of artifacts. Furthermore, the optimization step, the reprojection step and the second reconstruction step can be done iteratively with a predetermined amount of iterations or until the resultant reconstructed image meets a desired image quality metric.

Reconstruction Method for Increasing Image Resolution of Computed Tomography Systems

UW–Madison researchers have developed an image reconstruction method that increases image resolution with no hardware modifications to the CT system. The acquired projection views are backprojected onto an image grid with a higher resolution than the supported detector element resolution. This image is reprojected to obtain the original attenuation data and additional “pseudo” attenuation data. Effectively, this samples the area of interest at a density greater than or equal to the original scan. The original attenuation data is combined with the pseudo attenuation data to reconstruct a higher resolution image than would have been achieved with only the original data.

Image Reconstruction Method for Computed Tomography and Magnetic Resonance Cardiac Imaging

A UW–Madison researcher has developed a new method for reconstructing highly undersampled images at specific cardiac phases for both X-ray computed tomography (CT) and magnetic resonance imaging (MRI). The method uses a highly constrained backprojection method and requires a composite image that is enhanced using the previously proposed method.

The highly constrained backprojection reconstruction method weights image pixels to increase the image quality at areas in which the composite image pixels intersect structures in the subject, instead of simply assuming the pixels should be weighted evenly like previous techniques. Increasing the quality of this composite image directly increases the reconstructed image quality. The composite image can be enhanced further by subtracting the stationary tissue that surrounds the heart.

Highly Constrained Backprojected Reconstruction (HYPR) for Computed Tomography Images

A UW–Madison researcher has developed a new method for reconstructing CT images using an improved backprojection method. The method uses a composite image and an assumption of an inhomogeneous backprojected signal to weight the distribution of the backprojected views to reconstruct images. This allows quality CT images to be reconstructed using either less data to reduce scan time or a lower X-ray dose to reduce patient risk.

The composite image can be obtained from either the CT scan or previous data to enhance undersampled data sets. The improved back projection method enables images to be acquired with a lower X-ray dose without significant loss in the image’s signal-to- noise ratio (SNR). This is done by reconstructing a set of low dose images with one single high SNR composite image.

Highly Constrained Image Reconstruction for Medical Imaging Applications

A UW–Madison researcher has developed a new method for reconstructing medical images from projection views of a subject. A backprojection technique is used that does not assume homogeneity in the backprojected signal. A composite image is reconstructed, and then this composite image is used to highly constrain the image reconstruction process to provide more image detail where needed.

This image reconstruction method reduces scan time and radiation dose, and provides higher resolution for time-resolved studies. Acquiring a highly sampled composite image will increases the signal-to-noise ratio (SNR) of the undersampled reconstructed images. This method can be used to improve the reconstruction of medical images.

Cone-Beam Filtered Backprojection Image Reconstruction Method for Short Trajectories

A UW-Madison researcher has developed a method for accurately reconstructing images from divergent beams of acquired image data. In this new cone-beam filtered back projection (FBP) reconstruction method, a shift-invariant FBP algorithm is applied to the arc scaning path. The algorithm filters the pre-weighted and differentiated cone-beam projection data along some image voxel-dependent eigen-directions, providing better image quality than the conventional Feldkamp algorithm. Another advantage is that this algorithm works in super-short scanning mode—the normal angular range of projections is not necessary to satisfy the so-called short scan condition.

Fan- and Cone-Beam Image Reconstruction Using Filtered Backprojection of Differentiated Projection

A UW-Madison researcher has developed an improved algorithm that provides better image quality than FDK-type algorithms. This new algorithm was derived from cone-beam geometries. It is part of an image reconstruction method that produces images from projection data by filtering the backprojection image of differentiated projection data. The algorithm works with many different imaging modalities that use symmetric or asymmetric two-dimensional fan beams or three-dimensional cone beams.

High-Speed Computed Tomography System Using a Spherical Anode for Improved Medical Imaging

UW–Madison researchers have developed an X-ray tomographic system that uses a virtual spherical anode to break free of conventional planar CT data acquisition by acquiring 3-D projections. Instead of measuring radially in one plane, data is measured radially in a 3-D starburst pattern. This process provides a number of benefits, including the ability to collect missing X-ray beam data that has left the intended plane of interest, flexibility in projection angle selection to minimize or measure tissue motion, and the ability to rapidly collect sparse projection sets for large volumes on a real-time basis.

The X-ray tomographic system would consist of a patient support, an X-ray source, a multi-element detector, and a controller. Next to the patient support is the X-ray source, which consists of an electron gun and a spherical anode in between the patient support and the gun. Opposite the anode from the patient support is the multi-element detector that receives the X-rays. The multi-element detector and the X-ray source communicate via the controller, which steers the electron gun across the spherical anode to acquire a series of projection sets over a range of latitudinal and longitudinal angles.

Correction of CT Images for Truncated or Incomplete Projections

A UW-Madison researcher has developed a data consistency condition for estimating missing or contaminated values from the fan-beam projections used in CT. The data consistency condition is used to calculate individual measurements in a missing, noisy or contaminated projection based on measurements from other, uncorrupted projections acquired during the scan. The corrupted projection data is then replaced with the estimated values and the image is reconstructed from the corrected projections.

Fourier Space Tomographic Image Reconstruction Method

UW-Madison researchers have developed a fundamentally new framework for image reconstruction in computed tomography (CT) that extends the parallel beam projection-slice theorem to fan beam and cone beam projections. Key to the invention is a generalized projection-slice theorem (GSPT) solved by the researchers. Using this new theorem, the invention can directly construct the Fourier space of an image object during data acquisition from fan or cone beam projections, without employing Radon space. Once the Fourier space of the image object has been built, the invention allows ready reconstruction of the image by using the inverse Fourier transform.

Contrast Agents That Improve GI Tract Opacification During Abdominal and Pelvic CT Scans

UW-Madison researchers have developed improved contrast agents for use during visualization of the gastrointestinal tract during CT scans. The contrast agents include an iso-osmotic contrast agent that preferably comprises polyethylene glycol (PEG) and electrolytes that make the PEG iso-osmotic. A positive contrast agent, such as an iodine- or barium-based contrast agent, may be added to the PEG.

Improved CT Image Reconstruction Method for Use with 1-D Detector Arrays

A UW–Madison researcher has developed an improved algorithm for reconstructing CT images from scans performed with a fan beam source and a 1-D detector array. A major advantage of this algorithm is that, unlike those employed by most commercially available CT systems, images can be accurately reconstructed by using fan beam data acquired from a scanning path with angular coverage less than the standard 180+ fan angle. This feature improves the temporal resolution of fan beam CT cardiac imaging. The algorithm contains a filtered back projection (FBP) structure that allows the use of fast Fourier transform (FFT) to accelerate the image reconstruction process. It also acquires the data needed to image a region of interest with shorter scan paths than are required by previous algorithms involving FBP, potentially lowering the X-ray dosage associated with CT scanning.