Alan H. Baydush

Feasibility of optimizing the dose distribution in lung tumors using fluorine‐18‐fluorodeoxyglucose positron emission tomography and single photon emission computed tomography guided dose prescriptions

The information provided by functional images may be used to guide radiotherapy planning by identifying regions that require higher radiation dose. In this work we investigate the dosimetric feasibility of delivering dose to lung tumors in proportion to the fluorine-18-fluorodeoxyglucose activity distribution from positron emission tomography (FDG-PET). The rationale for delivering dose in proportion to the tumor FDG-PET activity distribution is based on studies showing that FDG uptake is correlated to tumor cell proliferation rate, which is shown to imply that this dose delivery strategy is theoretically capable of providing the same duration of local control at all voxels in tumor. Target dose delivery was constrained by single photon emission computed tomography (SPECT) maps of normal lung perfusion, which restricted irradiation of highly perfused lung and imposed dose-function constraints. Dose-volume constraints were imposed on all other critical structures. All dose-volume/function constraints were considered to be soft, i.e., critical structure doses corresponding to volume/function constraint levels were minimized while satisfying the target prescription, thus permitting critical structure doses to minimally exceed dose constraint levels. An intensity modulation optimization methodology was developed to deliver this radiation, and applied to two lung cancer patients. Dosimetric feasibility was assessed by comparing spatially normalized dose-volume histograms from the nonuniform dose prescription (FDG-PET proportional) to those from a uniform dose prescription with equivalent tumor integral dose. In both patients, the optimization was capable of delivering the nonuniform target prescription with the same ease as the uniform target prescription, despite SPECT restrictions that effectively diverted dose from high to low perfused normal lung. In one patient, both prescriptions incurred similar critical structure dosages, below dose-volume/function limits. However, in the other patient, critical structure dosage from the nonuniform dose prescription exceeded dose-volume/function limits, and greatly exceeded that from the uniform dose prescription. Strict compliance to dose-volume/ function limits would entail reducing dose proportionality to the FDG-PET activity distribution, thereby theoretically reducing the duration of local control. Thus, even though it appears feasible to tailor lung tumor dose to the FDG-PET activity distribution, despite SPECT restrictions, strict adherence to dose-volume/function limits could compromise the effectiveness of functional image guided radiotherapy.

Incorporation of an iterative, linear segmentation routine into a mammographic mass CAD system.

In previous research, we have developed a computer-aided detection (CAD) system designed to detect masses in mammograms. The previous version of our system employed a simple but imprecise method to localize the masses. In this research, we present a more robust segmentation routine for use with mammographic masses. Our hypothesis is that by more accurately describing the morphology of the masses, we can improve the CAD systems ability to distinguish masses from other mammographic structures. To test this hypothesis, we incorporated the new segmentation routine into our CAD system and examined the change in performance. The developed iterative, linear segmentation routine is a gray level-based procedure. Using the identified regions from the previous CAD system as the initial seeds, the new segmentation algorithm refines the suspicious mass borders by making estimates of the interior and exterior pixels. These estimates are then passed to a linear discriminant, which determines the optimal threshold between the interior and exterior pixels. After applying the threshold and identifying the objects outline, two constraints on the border are applied to reduce the influence of background noise. After the border is constrained, the process repeats until a stopping criterion is reached. The segmentation routine was tested on a study database of 183 mammographic images extracted from the Digital Database for Screening Mammography. Eighty-three of the images contained 50 malignant and 50 benign masses; 100 images contained no masses. The previously developed CAD system was used to locate a set of suspicious regions of interest (ROIs) within the images. To assess the performance of the segmentation algorithm, a set of 20 features was measured from the suspicious regions before and after the application of the developed segmentation routine. Receiver operating characteristic (ROC) analysis was employed on the ROIs to examine the discriminatory capabilities of each individual feature before and after the segmentation routine. A statistically significant performance increase was found in many of the individual features, particularly those describing the mass borders. To examine how the incorporation of the segmentation routine affected the performance of the overall CAD system, free-response ROC (FROC) analysis was employed. When considering only malignant masses, the FROC performance of the system with the segmentation routine appeared better than the previous system. When detecting 90% of the malignant masses, the previous system achieved 4.9 false positives per image (FPpI) compared to the post-segmentation systems 4.2 FPpI. At 80% sensitivity, the respective FPpI were 3.5 and 1.6.

Predicting radiotherapy-induced cardiac perfusion defects.

The purpose of this work is to compare the efficacy of mathematical models in predicting the occurrence of radiotherapy-induced left ventricular perfusion defects assessed using single-photon emission computed tomography (SPECT). The basis of this study is data from 73 left-sided breast/ chestwall patients treated with tangential photon fields. The mathematical models compared were three commonly used parametric models [Lyman normal tissue complication probability (LNTCP), relative serialty (RS), generalized equivalent uniform dose (gEUD)] and a nonparametric model (Linear discriminant analysis--LDA). Data used by the models were the left ventricular dose--volume histograms, or SPECT-based dose-function histograms, and the presence/absence of SPECT perfusion defects 6 months postradiation therapy (21 patients developed defects). For the parametric models, maximum likelihood estimation and F-tests were used to fit the model parameters. The nonparametric LDA model step-wise selected features (volumes/function above dose levels) using a method based on receiver operating characteristics (ROC) analysis to best separate the groups with and without defects. Optimistic (upper bound) and pessimistic (lower bound) estimates of each models predictive capability were generated using ROC curves. A higher area under the ROC curve indicates a more accurate model (a model that is always accurate has area = 1). The areas under these curves for different models were used to statistically test for differences between them. Pessimistic estimates of areas under the ROC curve using dose-volume histogram/ dose-function histogram inputs, in order of increasing prediction accuracy, were LNTCP (0.79/0.75), RS (0.80/0.77), gEUD (0.81/0.78), and LDA (0.84/0.86). Only the LDA model benefited from SPECT-based regional functional information. In general, the LDA model was statistically superior to the parametric models. The LDA model selected as features the left ventricular volumes above approximately 23 Gy (V23), essentially volume in field, and 33 Gy (V33), as best separating the groups with and without defects. In conclusion, the nonparametric LDA model appears to be a more accurate predictor of radiotherapy-induced left ventricular perfusion defects than commonly used parametric models.

Improved image quality in digital mammography with image processing

PURPOSE The effect of image processing, specifically Bayesian image estimation (BIE), on digital mammographic images is studied. BIE is an iterative, nonlinear statistical estimation technique that has previously been used in chest radiography to reduce image scatter content and improve the contrast-to-noise ratio (CNR). We adapt this technique to digital mammography and examine its effect. METHODS/MATERIALS Images of the American College of Radiologists (ACR) breast phantom were acquired on a calibrated digital mammography system at a normal mammographic exposure both with and without a grid. An iterative Bayesian estimation algorithm was formulated and used to process the images acquired without a grid. Quantitative scatter fractions were measured and compared for the image acquired with the grid, the image acquired without the grid, and the image acquired without the grid and processed by the Bayesian algorithm. CNR values were also computed for the four visible masses in the ACR phantom before and after processing and compared to a grid. RESULTS Initial images acquired without an antiscatter grid had scatter fractions of 0.46. Processing this image with BIE reduced the scatter content to under 0.04. In comparison, the image acquired with a grid had scatter of 0.19. BIE processing accounted for CNR improvements from 29% to 219% for the masses seen in the ACR phantom as compared to the unprocessed image. Visibility of the four masses in the phantom was improved. CONCLUSIONS Bayesian image estimation can be used with digital mammography to reduce scatter fractions. This technique is very useful as it can reduce scatter content effectively without introducing any adverse effects, such as grid line aliasing. Bayesian processing can also increase image CNR, which may potentially increase the visualization of subtle masses. Preliminary work shows an improvement in CNR to values greater than that provided by a standard grid.

SCATTER COMPENSATION FOR DIGITAL CHEST RADIOGRAPHY USING MAXIMUM LIKELIHOOD EXPECTATION MAXIMIZATION

RATIONALE AND OBJECTIVES.An iterative maximum likelihood expectation maximization algorithm (MLEM) has been developed for scatter compensation in chest radiography. METHODS.The MLEM technique produces a scatterreduced image which maximizes the probability of observing the measured image. We examined the scatter content and the low-contrast signal-to-noise ratio (SNR) in digital radiographs of anatomical phantoms before and after compensation. RESULTS.MLEM converged to an accurate (6.4%RMS residual scatter error) estimate within 12 iterations. Both contrast and noise were increased in the processed images as iteration progressed. In the lung, contrast was increased 108% and SNR was improved by 10%. In the retrocardiac region, contrast was increased 180% while SNR decreased by 6%. CONCLUSIONS.This is the first report of a post-acquisition scatter compensation technique which can increase SNR. These results suggest that statistical estimation techniques can enhance image quality and quantitative accuracy for digital chest radiography.

Initial application of digital tomosynthesis with on-board imaging in radiation oncology

We present preliminary investigations that examine the feasibility of incorporating digital tomosynthesis into radiation oncology practice with the use of kilovoltage on-board imagers (OBI). Modern radiation oncology linear accelerators now include hardware options for the addition of OBI for on-line patient setup verification. These systems include an x-ray tube and detector mounted directly on the accelerator gantry that rotate with the same isocenter. Applications include cone beam computed tomography (CBCT), fluoroscopy, and radiographs to examine daily patient positioning to determine if the patient is in the same location as the treatment plan. While CBCT provides the greatest anatomical detail, this approach is limited by long acquisition and reconstruction times and higher patient dose. We propose to examine the use of tomosynthesis reconstructed volumetric data from limited angle projection images for short imaging time and reduced patient dose. Initial data uses 61 projection images acquired over an isocentric arc of twenty degrees with the detector approximately fifty-four centimeters from isocenter. A modified filtered back projection technique, which included a mathematical correction for isocentric motion, was used to reconstruct volume images. These images will be visually and mathematically compared to volumetric computed tomography images to determine efficacy of this system for daily patient positioning verification. Initial images using the tomosynthesis reconstruction technique show much promise and bode well for effective daily patient positioning verification with reduced patient dose and imaging time. Additionally, the fast image acquisition may allow for a single breath hold imaging sequence, which will have no breath motion.

Improved Bayesian image estimation for digital chest radiography.

PURPOSE Previously, we have shown that Spatially Varying Bayesian Image Estimation (SVBIE) can be used to reduce scatter and improve contrast-to-noise ratios (CNR) in digital chest radiographs with no degradation of image resolution. This previous algorithm used a model for scatter compensation that was derived for emission tomography. Here, we develop and evaluate a new iterative SVBIE technique that incorporates a scatter model derived for projection radiography. MATERIALS AND METHODS Portable digital radiographs of an anthropomorphic chest phantom were obtained along with quantitative scatter measurements using a calibrated photostimulable phosphor system. The new iterative SVBIE technique was applied to the phantom image to reduce scatter. Scatter fraction reduction, CNR improvement, and resolution degradation were evaluated. RESULTS Residual scatter fractions were reduced to less than 2% in the lungs and 30% in the mediastinum at 14 iterations. CNR was improved by approximately 50% in the lung region and 187% in the mediastinum. Resolution was not degraded. CONCLUSIONS The new SVBIE technique can reduce scatter to levels far below those provided by an antiscatter grid and can increase CNR without loss of resolution. The new technique outperforms the previous Bayesian techniques.

Bi-plane correlation imaging for improved detection of lung nodules

Bi-plane correlation imaging (BCI) is a new imaging approach that utilizes angular information from a bi-plane digital acquisition in conjunction with computer assisted detection (CAD) to reduce the degrading influence of anatomical noise in the detection of subtle lesions in planar images. An anthropomorphic chest phantom, supplemented with added nodule phantoms (5-13 mm at the image plane), was imaged from different posterior projections within a ±12° range by moving the x-ray tube vertically and horizontally with respect to the detector. Each image was analyzed using a basic front-end single-view CAD algorithm. The correlation of the suspect lesions from the PA view with those from each of the oblique views was examined using a priori knowledge of the acquisition geometry. The correlated suspect lesions were registered as positive. Using an optimum --3° vertical geometry and processing parameters, BCI resulted in 62.5% sensitivity, 1.5 FP/image, and 0.885 PPV. The corresponding values from the observer experiment were 56% sensitivity, 10.8 FP/image, and 0.45 PPV, respectively. Compared to single-view CAD results, the BCI reduced sensitivity by 20%. However, the corresponding reduction in FPs was notably higher (94%) leading to 140% improvement in the PPV. Changes in processing parameters could result in higher PPV and lower FP/image at the expense of lower sensitivity. Similar findings were indicated for small (5-9 mm) and large (9-13 mm) nodules, but the relative improvement was significantly higher for smaller nodules. (The research was supported by a grant from the NIH, R21CA91806.)

A mammographic mass CAD system incorporating features from shape, fractal, and channelized Hotelling observer measurements: preliminary results

In this paper, we present preliminary results from a highly sensitive and specific CAD system for mammographic masses. For false positive reduction, the system incorporated features derived from shape, fractal, and channelized Hotelling observer (CHO) measurements. The database for this study consisted of 80 craniocaudal mammograms randomly extracted from USFs digital database for screening mammography. The database contained 49 mass findings (24 malignant, 25 benign). To detect initial mass candidates, a difference of Gaussians (DOG) filter was applied through normalized cross correlation. Suspicious regions were localized in the filtered images via multi-level thresholding. Features extracted from the regions included shape, fractal dimension, and the output from a Laguerre-Gauss (LG) CHO. Influential features were identified via feature selection techniques. The regions were classified with a linear classifier using leave-one-out training/testing. The DOG filter achieved a sensitivity of 88% (23/24 malignant, 20/25 benign). Using the selected features, the false positives per image dropped from ~20 to ~5 with no loss in sensitivity. This preliminary investigation of combining multi-level thresholded DOG-filtered images with shape, fractal, and LG-CHO features shows great promise as a mass detector. Future work will include the addition of more texture and mass-boundary descriptive features as well as further exploration of the LG-CHO.

Human-observer templates for detection of a simulated lesion in mammographic images

We describe a probit regression approach for maximum-likelihood (ML) estimation of a linear observer template from human-observer data in two-alternative forced-choice experiments. Like a previous approach to ML estimation in this problem [Abbey & Eckstein, Proc. SPIE, Vol. 4324, 2001], our approach does not make any assumptions about the distribution of the images. The previous approach utilized a regularizing prior distribution to control the degrees of freedom in the problem. In this work, we constrain the observer template to be represented by a limited number of linear features. Standard methods of probit regression are described for estimating the feature weights, and hence the observer templates. We have used this probit regression method to estimate human-observer templates for the detection of a small (5mm diameter) round simulated mass embedded in digitized mammograms. Our estimated templates for detecting the mass contain a band of heavily weighted spatial frequencies from 0.08 to 0.3 cycles/mm. We show comparisons between the human-observer template data, and the templates of a number of linear model observers that have been investigated as perceptual models of the human.