We evaluate a computer-aided diagnosis (CADx) system developed for both melanocytic and non-melanocytic skin lesions by using conventional digital photographs with lesion boundaries manually marked by a dermatologist. Clinical images of skin lesions taken by conventional digital cameras can capture useful information such as shape, color, and texture for diagnosing skin cancer. However, shape/border features are difficult to analyze automatically because skin surface reflections may change skin color and make segmentation a challenging task. In this study, two non-medical users manually mark the boundaries of a dataset of 769 (174 malignant, 595 benign) conventional photographs of melanocytic and non-melanocytic skin lesions. A state-of-the-art software system for segmenting color images, JSEG, is also tested on the same dataset. Their results are compared to a dermatologist's markings, which are used as the gold standard in this study. The human users' markings are relatively close to the gold standard and achieve an overlapping rate of 70.4% (+/- 15.3%, std) and 74.5% (+/- 14.7%, std). Compared to human users, JSEG only succeeds in segmenting 636 (82.7%) out of 769 lesions and achieves an overlapping rate of 72.4% (+/-20.4%) for these 636 lesions. The estimated area under the receiver operating characteristic curve (AUC) of the CADx by using lesion boundary markings of users 1, 2, and JSEG are 0.915, 0.940, and 0.857 respectively. Our preliminary results indicate that manual segmentation can be repeated relatively consistent compared to automatic segmentation.
In this paper, we present a texture analysis based method for diagnosing the Basal Cell Carcinoma (BCC) skin cancer
using optical images taken from the suspicious skin regions. We first extracted the Run Length Matrix and Haralick
texture features from the images and used a feature selection algorithm to identify the most effective feature set for the
diagnosis. We then utilized a Multi-Layer Perceptron (MLP) classifier to classify the images to BCC or normal cases.
Experiments showed that detecting BCC cancer based on optical images is feasible. The best sensitivity and specificity
we achieved on our data set were 94% and 95%, respectively.
Bony structures at the skull base were the main obstacle to detection and estimation of arterial stenoses and aneurysms
for CT angiography in the brain. Direct subtraction and the matched mask bone elimination (MMBE) have become two
standard methods for removing bony structures. However, clinicians regularly find that calcified plaques at or near the
carotid canal cannot be removed satisfactorily by existing methods. The blood-plaque boundary tends to be blurred by
subtraction operation while plaque size is constantly overestimated by the bone mask dilation operation in the MMBE
approach. In this study, we propose using the level of enhancement to adjust the MMBE bone mask more intelligently
on the artery- and tissue-bone/plaque boundaries. The original MMBE method is only applied to the tissue-bone
boundary voxels; while the artery-bone/blood-plaque boundary voxels, identified by a higher enhancement level, are
processed by direct subtraction instead. A dataset of 6 patients (3 scanned with a regular dose and 3 scanned with a
reduced dose) with calcified plaques at or near the skull base is used to examine our new method. Preliminary results
indicate that the visualization of intracranial arteries with calcified plaques at the skull base can be improved effectively
We evaluated a Pareto front-based multi-objective evolutionary algorithm for optimizing our CT colonography
(CTC) computer-aided detection (CAD) system. The system identifies colonic polyps based on curvature and
volumetric based features, where a set of thresholds for these features was optimized by the evolutionary algorithm.
We utilized a two-fold cross-validation (CV) method to test if the optimized thresholds can be generalized
to new data sets. We performed the CV method on 133 patients; each patient had a prone and a supine scan.
There were 103 colonoscopically confirmed polyps resulting in 188 positive detections in CTC reading from either
the prone or the supine scan or both. In the two-fold CV, we randomly divided the 133 patients into two
cohorts. Each cohort was used to obtain the Pareto front by a multi-objective genetic algorithm, where a set of
optimized thresholds was applied on the test cohort to get test results. This process was repeated twice so that
each cohort was used in the training and testing process once. We averaged the two training Pareto fronts as
our final training Pareto front and averaged the test results from the two runs in the CV as our final test results.
Our experiments demonstrated that the averaged testing results were close to the mean Pareto front determined
from the training process. We conclude that the Pareto front-based algorithm appears to be generalizable to
new test data.
We have found greater difficulty achieving desirable sensitivities and specificities with our computer-aided detection
(CAD) system on polyps sized 6-9 mm. Missed polyps in our ground truth CAD training datasets could be one possible
cause. Most CT colonography (CTC) protocols require supine and prone scans therefore the number of polyps visible to
a radiologist in at least one scan may increase. However, registration of a specific polyp visible in both scans can prove
difficult without a uniform coordinate system. Using a teniae coli registration tool we hypothesized we could register
and find a statistically significant number of 6-9 mm polyps believed to be not findable in one scan subsequently
reducing error in the training data and enabling better training of our CAD system. Database queries yielded 20 polyps
initially believed to be not findable in one scan. The teniae coli navigation and registration system allowed us to identify
30% (6/20) of the polyps as matches with confidence in both scans (rating 1) and 10% (2/20) of the polyps with a
potential match with some uncertainty (rating 2). No convincing match was found for 60% (12/20) of polyps (rating 3).
We conclude that this teniae coli registration tool is an effective means of identifying and reducing ground truth data
errors in 6-9 mm polyps initially believed not findable in one scan. The use of this tool has the potential to improve the
performance of a CAD system on the more difficult 6-9 mm polyps.
We evaluate and improve an existing curvature-based region growing algorithm for colonic polyp detection for our CT colonography (CTC) computer-aided detection (CAD) system by using Pareto fronts. The performance of a polyp detection algorithm involves two conflicting objectives, minimizing both false negative (FN) and false positive (FP) detection rates. This problem does not produce a single optimal solution but a set of solutions known as a Pareto front. Any solution in a Pareto front can only outperform other solutions in one of the two competing objectives. Using evolutionary algorithms to find the Pareto fronts for multi-objective optimization problems has been common practice for years. However, they are rarely investigated in any CTC CAD system because the computation cost is inherently expensive. To circumvent this problem, we have developed a parallel program implemented on a Linux cluster environment. A data set of 56 CTC colon surfaces with 87 proven positive detections of polyps sized 4 to 60 mm is used to evaluate an existing one-step, and derive a new two-step region growing algorithm. We use a popular algorithm, the Strength Pareto Evolutionary Algorithm (SPEA2), to find the Pareto fronts. The performance differences are evaluated using a statistical approach. The new algorithm outperforms the old one in 81.6% of the sampled Pareto fronts from 20 simulations. When operated at a suitable sensitivity level such as 90.8% (79/87) or 88.5% (77/87), the FP rate is decreased by 24.4% or 45.8% respectively.
Colonic polyps appear like elliptical protrusions on the inner wall of the colon. Curvature based features for colonic polyp detection have proved to be successful in several computer-aided diagnostic CT colonography (CTC) systems. Some simple thresholds are set for those features for creating initial polyp candidates, sophisticated classification scheme are then applied on these polyp candidates to reduce false positives. There are two objective functions, the number of missed polyps and false positive rate, that need to be minimized when setting those thresholds. These two objectives conflict and it is usually difficult to optimize them both by a gradient search. In this paper, we utilized a multiobjective evolutionary method, the Strength Pareto Evolutionary Algorithm (SPEA2), to optimize those thresholds. SPEA2 incorporates the concept of Pareto dominance and applies genetic techniques to evolve individual solutions to the Pareto front. The SPEA2 algorithm was applied to colon CT images from 27 patients each having a prone and a supine scan. There are 40 colonoscopically confirmed polyps resulting in 72 positive detections in CTC reading. The results obtained by SPEA2 were compared with those obtained by our old system, where an appropriate value was set for each of those thresholds by a histogram examination method. If we keep the sensitivity the same as that of our old system, the SPEA2 algorithm reduced false positive rate by 76.4% from average false positive 55.6 to 13.3 per data set. If the false positive rate is kept the same for both systems, SPEA2 increased the sensitivity by 13.1% from 53 to 61 among 72 ground truth detections.
We present a synchronous navigation module for CT colonography (CTC) reading. The need for such a system arises because most CTC protocols require a patient to be scanned in both supine and prone positions to increase sensitivity in detecting colonic polyps. However, existing clinical practices are limited to reading one scan at a time. Such limitation is due to the fact that building a reference system between scans for the highly flexible colon is a nontrivial task. The conventional centerline approach, generating only the longitudinal distance along the colon, falls short in providing the necessary orientation information to synchronize the virtual navigation cameras in both scanned positions. In this paper we describe a synchronous navigation system by using the teniae coli as anatomical references. Teniae coli are three parallel bands of longitudinal smooth muscle on the surface of the colon. They are morphologically distinguishable and form a piecewise triple helix structure from the appendix to the sigmoid colon. Because of these characteristics, they are ideal references to synchronize virtual cameras in both scanned positions. Our new navigation system consists of two side-by-side virtual colonoscopic view panels (for the supine and prone data sets respectively) and one single camera control unit (which controls both the supine and prone virtual cameras). The capability to examine the same colonic region simultaneously in both scanned images can raise an observer's confidence in polyp identification and potentially improve the performance of CT colonography.
The computed tomographic colonography (CTC) computer aided detection (CAD) program is a new method in development to detect colon polyps in virtual colonoscopy. While high sensitivity is consistently achieved, additional features are desired to increase specificity. In this paper, a wavelet analysis was applied to CTCCAD outputs in an attempt to filter out false positive detections.
52 CTCCAD detection images were obtained using a screen capture application. 26 of these images were real polyps, confirmed by optical colonoscopy and 26 were false positive detections. A discrete wavelet transform of each image was computed with the MATLAB wavelet toolbox using the Haar wavelet at levels 1-5 in the horizontal, vertical and diagonal directions. From the resulting wavelet coefficients at levels 1-3 for all directions, a 72 feature vector was obtained for each image, consisting of descriptive statistics such as mean, variance, skew, and kurtosis at each level and orientation, as well as error statistics based on a linear predictor of neighboring wavelet coefficients. The vectors for each of the 52 images were then run through a support vector machine (SVM) classifier using ten-fold cross-validation training to determine its efficiency in distinguishing polyps from false positives.
The SVM results showed 100% sensitivity and 51% specificity in correctly identifying the status of detections. If this technique were added to the filtering process of the CTCCAD polyp detection scheme, the number of false positive results could be reduced significantly.
Given a segmented CT scan data of the colon represented as a triangle mesh, our water-plane algorithm will detect polyp candidates. The water-plane method comprises of pouring water into a polyp protrusion from the outside of the colon and in raising the “water-plane” until it cannot be incremented any further without causing water leakage. The method starts at a vertex and uses average normal of all triangles adjacent to the starting vertex to generate the initial water-plane, which will make the starting vertex “wet” but leave its neighboring vertices “dry”. The method will continue to wet neighboring vertices one by one and then their neighbors and so on until the water-plane cannot move any further without causing water leakage. The water-plane movement alternates between just raising the water level in completely convex regions and tilting about one or two anchor vertices that have neighbors that would get wet if the water level was raised any more. The final set of wet vertices is a cluster that is an initial polyp candidate. The water-plane method was compared against the current polyp candidate detection method in our Computer Aided Detection of Colon Polyps software pipeline, called the surface curvature method. It finds clusters of connected vertices that all exhibit elliptical curvature. The water-plane method showed multiple improvements in polyp candidate detection. It detected polyp candidates missed by the surface curvature method. It exhibited continuous polyp candidate regions instead of non-uniform or incomplete regions detected by the surface curvature method. And finally, it avoided some false positive detections reported by surface curvature method.
Colonic polyps are growths on the inner wall of the colon. They appear like elliptical protrusions which can be detected by curvature-derived shape discriminators. For reasons of computation efficiency, much of the past work in computer-aided diagnostic CT colonography adopted kernel-based convolution methods in curvature estimation. However, kernel methods can yield erroneous results at thin structures where the gradient diminishes. In this paper, we investigate three surface patch fitting methods: Cubic B-spline, paraboloid, and quadratic polynomials. This "patch" approach is based on the fact that a surface can be re-oriented such that it can be approximated by a bivariate function locally. These patch methods are evaluated by synthesized data with various orientations and sampling sizes. We find that the cubic spline method performs best regardless of large orientation variances. Cubic spline and quadratic polynomial methods perform equally well for large samples while the latter performs better for small ones. Based on the performance evaluation, we propose a new, two-stage curvature estimation method. The cubic spline fitting is performed first for its insensitivity to orientation. If the spline fitting errs by more than a preset value (indicating high surface tortuosity), a small data sample is fitted by a quadratic function. The evaluation is performed on 29 patients (58 data sets). With 88.7% sensitivity, the average number of false positives per data set is reduced by 44.5% from 33.5 (kernel method) to 18.6 (new method).
Line and net patterns in a noisy environment exist in many biomedical images. Examples include: Blood vessels in angiography, white matter in brain MRI scans, and cell spindle fibers in confocal microscopic data. These piecewise linear patterns with a Gaussian-like profile can be differentiated from others by their distinctive shape characteristics. A shape-based modeling method is developed to enhance and segment line and net patterns. The algorithm is implemented in an enhancement/thresholding type of edge operators. Line and net features are enhanced by second partial derivatives and segmented by thresholding. The method is tested on synthetic, angiography, MRI, and confocal microscopic data. The results are compared to the implementation of matched filters and crest lines. It shows that our new method is robust and suitable for different types of data in a broad range of noise levels.
Photostructurable glass-ceramic materials have received significant attention due to their utility in aerospace engineering and micro technology. For example, the ability to fabricate structures in glass is important in the design and integration of micro scale electronic, optical and fluidic devices. Direct-write pulsed UV laser processing techniques have been utilized recently to create patterned 3D microstructures in a lithium-aluminosilicate glass. The direct-write microfabrication process involves the formation of an initial latent image in the glass via UV laser radiation. Thermal-induced ceramization is utilized to develop the latent image into a permanent image. Material removal and microstructure fabrication are then accomplished by preferential isotropic etching of the developed regions.
A pulsed UV laser volumetric direct-write patterning technique has been used to fabricate the structural members and key fluidic distribution systems of a miniature 100 gm mass spacecraft called the Co-Orbital Satellite Assistant (COSA). A photostructurable glass ceramic material enables this photo-fabrication process. The COSA is a miniature space vehicle designed to assist its host ship by serving as a maneuverable external viewing platform. Using orbital dynamics simulation software, a minimum (Delta) V solution has been found that allows a COSA vehicle to eject from the host and maneuver into an observation orbit about the host vehicle. The result of the simulant show that a cold gas propulsion system can adequately support the mission given a total fuel volume of 5 cm3. A prototype COSA with dimensions of 50 X 50 X 50 mm has been fabricated and assembled for simulation experiments on an air table. The vehicle is fashioned out of 7 laser patterned wafers, electronics boards and a battery. The patterned wafers include an integrated 2-axis propulsion system, a fuel tank and a propellant distribution system. The electronics portion of the COSA vehicle includes a wireless communication system, 2 microcontrollers for system, 2 microcontrollers for system control and MEMS gyros for relative attitude determination. The COSA vehicle is designed to be mass producible and scalable.
Miniaturization technologies such as Micro-Electro-Mechanical Systems (MEMS) have been used to fabricate a prototype 100-gm class cold gas propulsion system suitable for use on a Co-Orbiting Satellite Assistant (COSA). The propulsion system is fabricated from bonded layers of photostructurable glass (Foturan glass; the design is based on fabricating integrated modular parts. Thus, the propulsion system is mass producible, expandable, expendable (low unit cost), and highly integrated.