Adam Huang

Computer-aided diagnosis of skin lesions using conventional digital photography: a reliability and feasibility study.

Background Computer-aided diagnosis (CADx) software that provides a second opinion has been widely used to assist physicians with various tasks. In dermatology, however, CADx has been mostly limited to melanoma or melanocytic skin cancer diagnosis. The frequency of non-melanocytic skin cancers and the accessibility of regular digital macrographs have raised interest in developing CADx for broader applications. Objectives To investigate the feasibility of using CADx to diagnose both melanocytic and non-melanocytic skin lesions based on conventional digital photographic images. Methods This study was approved by an institutional review board, and the requirement to obtain informed consent was waived. In total, 769 conventional photographs of melanocytic and non-melanocytic skin lesions were retrospectively reviewed and used to develop a CADx system. Conventional and new color-related image features were developed to classify the lesions as benign or malignant using support vector machines (SVMs). The performance of CADx was compared with that of dermatologists. Results The clinicians overall sensitivity, specificity, and accuracy were 83.33%, 85.88%, and 85.31%, respectively. New color correlation and principal component analysis (PCA) features improved the classification ability of the baseline CADx (pu200a=u200a0.001). The estimated area under the receiver operating characteristic (ROC) curve (Az) of the proposed CADx system was 0.949, with a sensitivity and specificity of 85.63% and 87.65%, respectively, and a maximum accuracy of 90.64%. Conclusions We have developed an effective CADx system to classify both melanocytic and non-melanocytic skin lesions using conventional digital macrographs. The systems performance was similar to that of dermatologists at our institute. Through improved feature extraction and SVM analysis, we found that conventional digital macrographs were feasible for providing useful information for CADx applications. The new color-related features significantly improved CADx applications for skin cancer.

On Concise 3-D Simple Point Characterizations: A Marching Cubes Paradigm

The centerlines of tubular structures are useful for medical image visualization and computer-aided diagnosis applications. They can be effectively extracted by using a thinning algorithm that erodes an object layer by layer until only a skeleton is left. An object point is ldquosimplerdquo and can be safely deleted only if the resultant image is topologically equivalent to the original. Numerous characterizations of 3-D simple points based on digital topology already exist. However, little work has been done in the context of marching cubes (MC). This paper reviews several concise 3-D simple point characterizations in a MC paradigm. By using the Euler characteristic and a few newly observed properties in the context of connectivity-consistent MC, we present concise and more self-explanatory proofs. We also present an efficient method for computing the Euler characteristic locally for MC surfaces. Performance evaluations on different implementations are conducted on synthetic data and multidetector computed tomography examination of virtual colonoscopy and angiography.

A robust hair segmentation and removal approach for clinical images of skin lesions

Artifacts such as hair are major obstacles to automatic segmentation of pigmented skin lesion images for computer-aided diagnosis systems. It is even more challenging to process clinical images taken by a regular digital camera, where the shadows of the skin texture may mimic hair-like curvilinear structures. In this study, we examined the popular DullRazor software with a dataset of 20 clinical images. The software, specifically designed for dermoscopic images, was unable to remove fine hairs or hairs in the shade. Alternatively, we proposed using conventional matched filters to enhance curvilinear structures. The more complicate hair intersection patterns, which were known to generate low matched filtering responses, were recovered by using region growing algorithms from nearby detected hair segments with linear discriminant analysis (LDA) based on a color similarity criterion. The preliminary results indicated the proposed method was able to remove more fine hairs and hairs in the shade, and lower false hair detection rate by 58% (from 0.438 to 0.183) as compared to the DullRazors approach.

CT Angiography Findings in Carotid Blowout Syndrome and Its Role as a Predictor of 1-Year Survival

BACKGROUND AND PURPOSE: Carotid blowout is a serious late complication of prior treatment of advanced head and neck cancer. We evaluate the efficacy of CTA in the diagnosis of impending carotid blowout syndrome in patients with head and neck cancer, and its capability to predict clinical outcome. MATERIALS AND METHODS: The clinical data of 29 patients with impending carotid blowout who underwent CTA were collected and analyzed. Imaging signs included tissue necrosis, exposed artery, viable perivascular tumor, pseudoaneurysm, and contrast extravasation. DSA was obtained in 20 patients. One-year outcomes were compared based on management. RESULTS: The most common CTA finding was necrosis (94%), followed by exposed artery (73%), viable tumor (67%), pseudoaneurysm (58%), and contrast extravasation (30%). Exposed artery, pseudoaneurysm, and contrast extravasation were the 3 CTA findings related to outcomes. All of the pseudoaneurysm and contrast extravasation cases were associated with an exposed artery. An exposed artery was the most important prognostic predictor and could not be diagnosed on DSA. Patients without the 3 findings on CTA (group 1) had the best survival rate at 1-year follow-up, followed by patients with the 3 findings treated immediately by permanent artery occlusion (group 2). Patients with the 3 findings who had no immediate treatment (group 3) had the worst outcomes (P < .001 in group 1 vs group 3 and group 2 vs group 3; P = .056 group 1 vs group 2). CONCLUSIONS: CTA, with its ability to diagnose an exposed artery compared with DSA, may offer important management and prognostic information in patients with impending carotid blowout.

Spherical parameterization of marching cubes IsoSurfaces based upon nearest neighbor coordinates

We present some new methods for parameterizing the triangle mesh surface (TMS) which result from the Marching Cubes (MC) algorithm. The methods apply to surfaces of genus zero and the parameter domain is a unit sphere. We take advantage of some special properties of the TMS resulting from the MC algorithm to obtain simple, computational efficient representations of the nearest neighbor coordinates and utilize these coordinates in the characterization of the parameterization by means of systems of equations which are solved iteratively. Examples and comparisons are presented.

Parameterizing marching cubes isosurfaces with natural neighbor coordinates

The triangular mesh surfaces (TMS) which result form the Marching Cubes (MC) algorithm have some unique and special properties not shared by general TMS. We exploit some of these properties in the development of some new, effective and efficient methods for parameterizing these surfaces. The parameterization consists of a planar triangulation which is isomorphic (maps one-to-one) to the triangular mesh. The parameterization is computed as the solution of a sparse linear system of equations which is based upon the fact that locally the MC surfaces are functions (height-fields). The coefficients of the linear system utilize natural neighbor coordinates (NNC) which depend upon Dirchlet tessellations. While the use of NNC for general TMS can be somewhat computationally expensive and is often done procedurally, for the present case of MC surfaces, we are able to obtain simple and explicit formulas which lead to efficient computational algorithms.

Using standard nonenhanced axial scans for cerebral CT angiography bone elimination: feasibility study.

Objective:To investigate the feasibility of using standard nonenhanced axial-mode scans as precontrast scans for bone elimination in cerebral CT angiography (CTA). Materials and Methods:A consecutive dataset of 32 patients who had both cerebral nonenhanced CT (NECT) (scanned in axial mode) and subtraction CTA (scanned in helical mode) examinations between April and August 2008 were retrospectively analyzed. For each patient, both axial- and helical-mode, NECT scans were processed by using the matched mask bone elimination (MMBE) method. Bone masks generated from axial- and helical-mode NECT scans were quantitatively compared by using overlapping analyses. The diagnostic quality and noise level of the resultant, maximum intensity projection, images by using 2 different bone masks were visually evaluated by 2 neuroradiologists independently using a 5-point scale (inferior, 1; worse, 2; equivalent, 3; better, 4; superior, 5). The effective doses to patients were estimated by using a dose-length product method. Results:Of the 28 (87.5%) patients without intrascan movements, overlap rates between axial- and helical-mode bone masks ranged from 99.2% to 99.9% (mean, 99.7% ± 0.2%). The mean diagnostic quality and noise level scores of resultant maximum intensity projection images given by 2 neuroradiologists were 3.0 ± 0.3 and 2.5 ± 0.5, respectively. The effective dose to patients with a routine brain CTA examination can be reduced from 1.16 to 0.78 mSv (16 cm, field-of-view) by using the proposed method if standard axial-mode NECT scans of the head are readily available. Conclusion:We found that using standard axial-mode NECT scans for bone elimination in helical-mode CTA is feasible. This method can further lower radiation dose without compromising the diagnostic quality.

Capillary detection for clinical images of basal cell carcinoma

Dilated capillaries are an important characteristic of basal cell carcinoma (BCC). Detecting capillaries in images can improve a computer-aided skin cancer diagnosis system. In this study, we investigate the feasibility to extract capillaries from clinical images of skin lesions recorded by a regular digital camera. First, we used a compact set of 1 curvilinear and 2 color parameters to train a support vector machine (SVM) classifier to identify capillary pixels. Second, the identified pixels were grouped by a region-growing algorithm to form capillary candidates. Last, the likelihood to be a true capillary was estimated based on the distance to the red color in the “CIE Lab” color space. The method was tested on a dataset of 21 BCC images with visible capillaries and 28 benign pigmented lesions without visible capillaries. The accuracy, sensitivity, and specificity of the proposed method were 89.8% (44/49), 90.5% (19/21), and 89.3% (25/28) respectively. We found capillaries recorded by a regular digital camera can be detected successfully.

The feasibility of using manual segmentation in a multifeature computer-aided diagnosis system for classification of skin lesions: a retrospective comparative study

Objectives To investigate the feasibility of manual segmentation by users of different backgrounds in a previously developed multifeature computer-aided diagnosis (CADx) system to classify melanocytic and non-melanocytic skin lesions based on conventional digital photographic images. Methods In total, 347 conventional photographs of melanocytic and non-melanocytic skin lesions were retrospectively reviewed, and manually segmented by two groups of physicians, dermatologists and general practitioners, as well as by an automated segmentation software program, JSEG. The performance of CADx based on inputs from these two groups of physicians and that of the JSEG program was compared using feature agreement analysis. Results The estimated area under the receiver operating characteristic curve for classification of benign or malignant skin lesions based were comparable on individual segmentation by the gold standard (0.893, 95% CI 0.856 to 0.930), dermatologists (0.886, 95% CI 0.863 to 0.908), general practitioners (0.883, 95% CI 0.864 to 0.903) and JSEG (0.856, 95% CI 0.812 to 0.899). The agreement in the malignancy probability scores among the physicians was excellent (intraclass correlation coefficient: 0.91). By selecting an optimal cut-off value of malignancy probability score, the sensitivity and specificity were 80.07% and 81.47% for dermatologists and 79.90% and 80.20% for general practitioners. Conclusions This study suggests that manual segmentation by general practitioners is feasible in the described CADx system for classifying benign and malignant skin lesions.

Evaluation of a computer-aided skin cancer diagnosis system for conventional digital photography with manual segmentation

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 dermatologists 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.