María J. Lado

Automatic detection of breast border and nipple in digital mammograms

Advances in the area of computerized image analysis applied to mammography may have very important practical applications in automatically detecting asymmetries (masses, architectural distortions, etc.) between the two breasts. We have developed a fully automatic technique to detect the breast border and the nipple, this being a necessary prerequisite for further image analysis. To detect the breast border, an algorithm that computes the gradient of gray levels was applied. To detect the nipple, three algorithms were compared (maximum height of the breast border, maximum gradient, and maximum second derivative of the gray levels across the median-top section of the breast). A combined method was also designed. The algorithms were tested on 156 digitized mammograms. The breast segmentation results were evaluated by two expert radiologists and one physicist. In 89% of the mammograms, the computed border was in close agreement with the radiologists estimated border. Segmentation results were acceptable to be used in computer-aided diagnostic schemes. The mean distance between the position of the nipple indicated by two radiologists by consensus and the position calculated by the computer was 6 mm.

Computer‐aided diagnosis: Automatic detection of malignant masses in digitized mammograms

A computerized method to automatically detect malignant masses on digital mammograms based on bilateral subtraction to identify asymmetries between left and right breast images was developed. After the digitization, in order to align left and right mammograms the breast border and nipple were automatically detected. Images were corrected to avoid differences in brightness due to the recording procedure. Left and right mammograms were subtracted and a threshold was applied to obtain a binary image with the information of suspicious areas. The suspicious regions or asymmetries were delimited by a region growing algorithm. Size and eccentricity tests were used to eliminate false-positive responses and texture features were extracted from suspicious regions to reject normal tissue regions. The scheme, tested in 70 pairs of digital mammograms, achieved a true-positive rate of 71% with an average number of 0.67 false positives per image. Computerized detection was evaluated by using free-response operating characteristic analysis (FROC). An area under the AFROC (A1) of 0.667 was obtained. Our results show that the scheme may be helpful to the radiologists by serving as a second reader in mammographic screening. The low number of false positives indicates that our scheme would not confuse the radiologist by suggesting normal regions as suspicious.

Image compression: Maxshift ROI encoding options in JPEG2000

Image compression can improve the performance of the digital systems by reducing time and cost in image storage and transmission without significant reduction of the image quality. Furthermore, the JPEG2000 has emerged as the new state-of-the art standard for image compression. In this paper, a Selective Coefficient Mask Shift (SCMShift) coding method is proposed. The technique, implemented over regions of interest (ROIs), is based on shifting the wavelet coefficients that belong to different subbands, depending on the coefficients relative to the original image. This method allows: (1) codification of multiple ROIs at various degrees of interest, (2) arbitrary shaped ROI coding, and (3) flexible adjustment of the compression quality of the ROI and the background. No standard modification for JPEG200 decoder was required. The method was applied over different types of images. Results show a better performance for the selected regions, when ROI coding methods were employed for the whole set of images. We believe that this method is an excellent tool for future image compression research, mainly on images where ROI coding can be of interest, such as the medical imaging modalities and several multimedia applications.

Application of the iris filter for automatic detection of pulmonary nodules on computed tomography images

We have developed a computer-aided diagnosis (CAD) system to detect pulmonary nodules on thin-slice helical computed tomography (CT) images. We have also investigated the capability of an iris filter to discriminate between nodules and false-positive findings. Suspicious regions were characterized with features based on the iris filter output, gray level and morphological features, extracted from the CT images. Functions calculated by linear discriminant analysis (LDA) were used to reduce the number of false-positives. The system was evaluated on CT scans containing 77 pulmonary nodules. The system was trained and evaluated using two completely independent data sets. Results for a test set, evaluated with free-response receiver operating characteristic (FROC) analysis, yielded a sensitivity of 80% at 7.7 false-positives per scan.

Detecting Sleep Apnea by Heart Rate Variability Analysis: Assessing the Validity of Databases and Algorithms

Obstructive sleep apnea (OSA) is a serious disorder caused by intermittent airway obstruction which may have dangerous impact on daily living activities. Heart rate variability (HRV) analysis could be used for diagnosing OSA, since this disease affects HRV during sleep. In order to validate different algorithms developed for detecting OSA employing HRV analysis, several public or proprietary data collections have been employed for different research groups. However, for validation purposes, it is obvious and evident the lack of a common standard database, worldwide recognized and accepted by the scientific community. In this paper, different algorithms employing HRV analysis were applied over diverse public and proprietary databases for detecting OSA, and the outcomes were validated in terms of a statistical analysis. Results indicate that the use of a specific database may strongly affect the performance of the algorithms, due to differences in methodologies of processing. Our results suggest that researchers must strongly take into consideration the database used when quoting their results, since selected cases are highly database dependent and would bias conclusions.

An open source tool for heart rate variability spectral analysis

In this paper we describe a software package for developing heart rate variability analysis. This package, called RHRV, is a third party extension for the open source statistical environment R, and can be freely downloaded from the R-CRAN repository. We review the state of the art of software related to the analysis of heart rate variability (HRV). Based upon this review, we motivate the development of an open source software platform which can be used for developing new algorithms for studying HRV or for performing clinical experiments. In particular, we show how the RHRV package greatly simplifies and accelerates the work of the computer scientist or medical specialist in the HRV field. We illustrate the utility of our package with practical examples.

Evaluation of an automated wavelet-based system dedicated to the detection of clustered microcalcifications in digital mammograms

Mammographic screening programs are delivering reductions in breast cancer mortality. However, breast cancer screening will be cost effective and will provide a real profit only when both high sensitivity and specificity levels are reached. To date, due to human or technical factors, a significant number of breast cancers are still missed or misinterpreted on the mammograms. Computer methodologies, developed to assist radiologists, could represent further amelioration by increasing diagnostic accuracy in the screening programs. We have tested a computerized scheme to detect clustered microcalcifications in digital mammograms, employing 360 mammograms that were randomly selected from the mammographic screening program, currently undergoing at the Galicia Community (Spain). After the digitization process, the breast border was initially determined. A wavelet-based algorithm was employed to detect the clusters of microcalcifications. The performance of the automated system over the test set was evaluated employing Free-response Receiver Operating Characteristic (FROC) methodology. The sensitivity achieved was 74% at a false positive detection rate of 1.83. The corresponding area under the Alternative FROC (AFROC) curve was A1=0.667 +/-0.09.Mammographic screening programs are delivering reductions in breast cancer mortality. However, breast cancer screening will be cost effective and will provide a real profit only when both high sensitivity and specificity levels are reached. To date, due to human or technical factors, a significant number of breast cancers are still missed or misinterpreted on the mammograms. Computer methodologies, developed to assist radiologists, could represent further amelioration by increasing diagnostic accuracy in the screening programs. We have tested a computerized scheme to detect clustered microcalcifications in digital mammograms, employing 360 mammograms that were randomly selected from the mammographic screening program, currently undergoing at the Galicia Community (Spain). After the digitization process, the breast border was initially determined. A wavelet-based algorithm was employed to detect the clusters of microcalcifications. The performance of the automated system over the test set was evaluated employing Free-response Receiver Operating Characteristic (FROC) methodology. The sensitivity achieved was 74% at a false positive detection rate of 1.83. The corresponding area under the Alternative FROC (AFROC) curve was A 1 =0.667 - 0.09.

Nocturnal evolution of heart rate variability indices in sleep apnea

Heart rate variability (HRV) is a valuable clinical tool in diagnosing multiple diseases. This paper presents the results of a spectral HRV analysis conducted with 46 patients. HRV indices for the whole night show differences among patients with severe and mild apnea, and healthy subjects. These differences also appear when performing the analysis over 5-min intervals, regarding apneas being present or not in the intervals. Differences were also observed when analyzing the HRV nocturnal evolution. Results are consistent with the hypothesis that cardiovascular risk remains constant for OSA patients while it increases towards the end of the night for healthy subjects.

Real and simulated clustered microcalcifications in digital mammograms. ROC study of observer performance.

We have developed a model to simulate clustered microcalcifications on digital mammograms. Wavelet transform techniques were used to detect real clustered microcalcifications. A feature analysis process was applied to automatically extract the features describing the individual simulated microcalcifications and clusters from the values of the real clustered microcalcifications present in the mammogram. Subsequently, a database of simulated and real clustered microcalcifications was created. Clusters of microcalcifications from this database were tested for indistinguishability from real ones. Two radiologists and one physicist were asked to indicate whether the microcalcifications were either real or simulated. The responses of the readers were evaluated with a ROC analysis and the area under the curve was calculated. The average ROC area was 0.54 +/- 0.03, indicating there was no statistical difference between real and simulated clustered microcalcifications. The method allows for the creations of simulated clustered microcalcifications that are virtually indistinguishable from real microcalcifications in digital mammograms and could be used to evaluate different image processing techniques.

Nonparametric estimation of conditional ROC curves: Application to discrimination tasks in computerized detection of early breast cancer

A local linear method for estimating the conditional ROC curve under the presence of continuous and categorical covariates is introduced. A data driven smoothing parameter selector based on the bootstrap is proposed. The methods are illustrated with real data from a discrimination problem emerging in the context of computer-aided diagnosis. The bootstrap approach is also used to construct pointwise confidence intervals for the area under the ROC curve.