Jorge Juan Suárez-Cuenca

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.

Quantification of Right and Left Ventricular Function in Cardiac MR Imaging: Comparison of Semiautomatic and Manual Segmentation Algorithms

The purpose of this study was to evaluate the performance of a semiautomatic segmentation method for the anatomical and functional assessment of both ventricles from cardiac cine magnetic resonance (MR) examinations, reducing user interaction to a “mouse-click”. Fifty-two patients with cardiovascular diseases were examined using a 1.5-T MR imaging unit. Several parameters of both ventricles, such as end-diastolic volume (EDV), end-systolic volume (ESV) and ejection fraction (EF), were quantified by an experienced operator using the conventional method based on manually-defined contours, as the standard of reference; and a novel semiautomatic segmentation method based on edge detection, iterative thresholding and region growing techniques, for evaluation purposes. No statistically significant differences were found between the two measurement values obtained for each parameter (p > 0.05). Correlation to estimate right ventricular function was good (r > 0.8) and turned out to be excellent (r > 0.9) for the left ventricle (LV). Bland-Altman plots revealed acceptable limits of agreement between the two methods (95%). Our study findings indicate that the proposed technique allows a fast and accurate assessment of both ventricles. However, further improvements are needed to equal results achieved for the right ventricle (RV) using the conventional methodology.

A hybrid method based on level set and 3D region growing for segmentation of the thoracic aorta

Abstract This study sought to develop a completely automatic method for image segmentation of the thoracic aorta. We used a total of 4682 images from 10 consecutive patients. The proposed method is based on the use of level set and region growing, automatically initialized using the Hough transform. The results obtained were compared to those of manual segmentation as performed by an external expert radiologist. Concordance between the developed method and manual segmentation ranged from 92.79 to 95.77% in the descending regions of the aorta and from 90.68 to 96.54% in the ascending regions, with a mean value of 93.83% being obtained for total segmentation.

Automatic characterization of thoracic aortic aneurysms from CT images

Accurate determination of the diameter is an important step for diagnosis and follow-up of aortic abnormalities such as aneurysms, caused by dilation of the vessel lumen. In this work we focus on the development of an automatic method for measuring the calibre of the thoracic aorta. The method is based on the application of principal component analysis on normal planes extracted from the aorta to establish the main axis of each section of the vessel. Two experiments were performed in order to test the accuracy and the rotational invariance of the developed method. Accuracy was determined by using a database of 15 clinical cases, where our method and a commercial software, which was considered as the gold standard, were compared. For the rotational invariance check, phantom images in different orientations were obtained and the diameter was measured with the proposed method. For clinical cases, a good agreement was observed between our method and the gold standard. The Bland Altman plots indicated that all of the values were within the acceptable limits of agreement with a bias of 0.2mm between both methods. For phantom cases, an ANOVA test revealed that the results achieved for the data sets acquired for the different orientations were not statistically different (F=1.88, p=0.153), which demonstrates the robustness of the method for rotations. The proposed method is applicable for measuring the diameter in all tested cases, and the results achieved underscored the capability of our approach for automatic characterization of thoracic aortic aneurysms.

INTEGRATION OF MULTIPLE CLASSIFIERS FOR COMPUTERIZED DETECTION OF LUNG NODULES IN CT

The purpose of this study was to investigate the usefulness of various classifier combination methods for improving the performance of a computer-aided diagnosis (CAD) system for pulmonary nodule detection in computed tomography (CT). We employed 85 CT scans with 110 nodules in the publicly available Lung Image Database Consortium (LIDC) dataset. We first applied our CAD scheme trained previously to the LIDC cases for identifying initial nodule candidates, and extracting 18 features for each nodule candidate. We used eight individual classifiers for false positives (FPs) reduction, including linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), Naive Bayes, simple logistic, artificial neural network (ANN) and support vector machines (SVMs) with three different kernels. Five classifier combination methods were then employed to integrate the outputs of the eight individual classifiers for improving detection performance. The five combination methods included two supervised (a likelihood ratio (LR) method and a probability method based on the output scores of the eight individual classifiers) and three unsupervised ones (the sum, the product and the majority voting of the output scores from the eight individual classifiers). Leave-one-case-out approach was employed to train and test individual classifiers and supervised combination methods. At a sensitivity of 80%, the numbers of FPs per CT scan for the eight individual classifiers were 6.1 for LDA, 19.9 for QDA, 10.8 for Naive Bayes, 8.4 for simple logistic, 8.6 for ANN, 23.7 for SVM-dot, 17.0 for SVM-poly, and 23.4 for SVM-anova; the numbers of FPs per CT scan for the five combination methods were 3.3 for the majority voting method, 5.0 for the sum, 4.6 for the product, 65.7 for the LR and 3.9 for the probability method. Compared to the best individual classifier, the majority voting method reduced 45% of FPs at 80% sensitivity. The performance of our CAD can be improved by combining multiple classifiers. The majority voting method achieved higher performance levels than other combination methods and all individual classifiers.

Integrating CAD modules in a PACS environment using a wide computing infrastructure

PurposeThe aim of this paper is to describe a project designed to achieve a total integration of different CAD algorithms into the PACS environment by using a wide computing infrastructure.MethodsThe aim is to build a system for the entire region of Galicia, Spain, to make CAD accessible to multiple hospitals by employing different PACSs and clinical workstations. The new CAD model seeks to connect different devices (CAD systems, acquisition modalities, workstations and PACS) by means of networking based on a platform that will offer different CAD services. This paper describes some aspects related to the health services of the region where the project was developed, CAD algorithms that were either employed or selected for inclusion in the project, and several technical aspects and results.ResultsWe have built a standard-based platform with which users can request a CAD service and receive the results in their local PACS. The process runs through a web interface that allows sending data to the different CAD services. A DICOM SR object is received with the results of the algorithms stored inside the original study in the proper folder with the original images.ConclusionsAs a result, a homogeneous service to the different hospitals of the region will be offered. End users will benefit from a homogeneous workflow and a standardised integration model to request and obtain results from CAD systems in any modality, not dependant on commercial integration models. This new solution will foster the deployment of these technologies in the entire region of Galicia.

Análisis de la función ventricular derecha en resonancia magnética cardíaca. Comparación de la cuantificación en los planos eje corto y 4 cámaras

OBJECTIVE To compare the methods of right ventricle segmentation in the short-axis and 4-chamber planes in cardiac magnetic resonance imaging and to correlate the findings with those of the tricuspid annular plane systolic excursion (TAPSE) method in echocardiography. MATERIAL AND METHODS We used a 1.5T MRI scanner to study 26 patients with diverse cardiovascular diseases. In all MRI studies, we obtained cine-mode images from the base to the apex in both the short-axis and 4-chamber planes using steady-state free precession sequences and 6mm thick slices. In all patients, we quantified the end-diastolic volume, end-systolic volume, and the ejection fraction of the right ventricle. On the same day as the cardiac magnetic resonance imaging study, 14 patients also underwent echocardiography with TAPSE calculation of right ventricular function. RESULTS No statistically significant differences were found in the volumes and function of the right ventricle calculated using the 2 segmentation methods. The correlation between the volume estimations by the two segmentation methods was excellent (r=0,95); the correlation for the ejection fraction was slightly lower (r=0,8). The correlation between the cardiac magnetic resonance imaging estimate of right ventricular ejection fraction and TAPSE was very low (r=0,2, P<.01). CONCLUSION Both ventricular segmentation methods quantify right ventricular function adequately. The correlation with the echocardiographic method is low.

AUTOMATIC DETECTION OF PULMONARY NODULES IN MDCT: PERFORMANCE EVALUATION WITH INDEPENDENT DATASETS

The purpose of this study was to evaluate the performance of a computer-aided diagnosis (CAD) system on the detection of pulmonary nodules in multidetector row computed tomography (MDCT) images by using two independent datasets. We collected CT cases of 63 patients with 132 nodules ranging 4–30 mm in diameter from a hospital in Spain (20 patients) and a hospital in France (43 patients). CT examinations were acquired by using a SOMATOM Emotion CT scanner in Spain, and a dual-source SOMATOM Definition CT scanner in France (Siemens Medical System, Forchheim, Germany), with the following parameters: 6 × 1.0 mm collimation, 130 kVp, 70 mA (Emotion 6); or 64 × 0.6 mm collimation, 100–120 kVp, and 100–110 mAs (Definition 64). Nodules were detected independently by three experienced chest radiologists, and their detection results were used as the reference standard. The CAD scheme was developed with an advanced 3D iris filter for improving nodule detection. The performance of the CAD scheme was tested with an independent evaluation method based on the two databases. Free-response receiver operating characteristic curves, sensitivity and number of false-positive per scan, were employed to evaluate the performance of the CAD scheme. The study was approved by the Institutional Review Board. At an average false positive (FP) rate of 5 per scan, our CAD scheme achieved sensitivities of 79.5% for all nodules, 80.3% for solid, 60.0% for non-solid, 58.1% for spiculated, and 86.1% for non-spiculated nodules. In conclusion, our CAD scheme could be utilized to help radiologists in the detection of lung nodules in CT. However, in this study we confirmed that significant differences could be found in the performance of the system depending on the testing database.

A CAD SCHEME FOR EARLY LUNG CANCER DETECTION IN CHEST RADIOGRAPHY

The purpose of this work is to describe a chest radiography computer-aided diagnostic (CAD) scheme designed to analyze the chest radiographs performed in the framework of the Galician (Spain) Healt...

Clinical Evaluation of an Automatic Method for Segmentation and Characterization of the Thoracic Aorta with and Without Aneurysm Patients

The aim of this study was to clinically evaluate a fully automatic method for segmentation and characterizing the thoracic aorta. From 2010 to 2013, a total of 27 patients were randomly selected for the study. The automatic method was compared with two segmentations manually performed by two independent radiologists amd a commercial software for gauging calibres.