Donato Cascio

Automatic detection of lung nodules in CT datasets based on stable 3D mass-spring models

We propose a computer-aided detection (CAD) system which can detect small-sized (from 3mm) pulmonary nodules in spiral CT scans. A pulmonary nodule is a small lesion in the lungs, round-shaped (parenchymal nodule) or worm-shaped (juxtapleural nodule). Both kinds of lesions have a radio-density greater than lung parenchyma, thus appearing white on the images. Lung nodules might indicate a lung cancer and their early stage detection arguably improves the patient survival rate. CT is considered to be the most accurate imaging modality for nodule detection. However, the large amount of data per examination makes the full analysis difficult, leading to omission of nodules by the radiologist. We developed an advanced computerized method for the automatic detection of internal and juxtapleural nodules on low-dose and thin-slice lung CT scan. This method consists of an initial selection of nodule candidates list, the segmentation of each candidate nodule and the classification of the features computed for each segmented nodule candidate.The presented CAD system is aimed to reduce the number of omissions and to decrease the radiologist scan examination time. Our system locates with the same scheme both internal and juxtapleural nodules. For a correct volume segmentation of the lung parenchyma, the system uses a Region Growing (RG) algorithm and an opening process for including the juxtapleural nodules. The segmentation and the extraction of the suspected nodular lesions from CT images by a lung CAD system constitutes a hard task. In order to solve this key problem, we use a new Stable 3D Mass-Spring Model (MSM) combined with a spline curves reconstruction process. Our model represents concurrently the characteristic gray value range, the directed contour information as well as shape knowledge, which leads to a much more robust and efficient segmentation process. For distinguishing the real nodules among nodule candidates, an additional classification step is applied; furthermore, a neural network is applied to reduce the false positives (FPs) after a double-threshold cut. The system performance was tested on a set of 84 scans made available by the Lung Image Database Consortium (LIDC) annotated by four expert radiologists. The detection rate of the system is 97% with 6.1 FPs/CT. A reduction to 2.5 FPs/CT is achieved at 88% sensitivity. We presented a new 3D segmentation technique for lung nodules in CT datasets, using deformable MSMs. The result is a efficient segmentation process able to converge, identifying the shape of the generic ROI, after a few iterations. Our suitable results show that the use of the 3D AC model and the feature analysis based FPs reduction process constitutes an accurate approach to the segmentation and the classification of lung nodules.

Fuzzy technique for microcalcifications clustering in digital mammograms

BackgroundMammography has established itself as the most efficient technique for the identification of the pathological breast lesions. Among the various types of lesions, microcalcifications are the most difficult to identify since they are quite small (0.1-1.0 mm) and often poorly contrasted against an images background. Within this context, the Computer Aided Detection (CAD) systems could turn out to be very useful in breast cancer control.MethodsIn this paper we present a potentially powerful microcalcifications cluster enhancement method applicable to digital mammograms. The segmentation phase employs a form filter, obtained from LoG filter, to overcome the dependence from target dimensions and to optimize the recognition efficiency. A clustering method, based on a Fuzzy C-means (FCM), has been developed. The described method, Fuzzy C-means with Features (FCM-WF), was tested on simulated clusters of microcalcifications, implying that the location of the cluster within the breast and the exact number of microcalcifications are known.The proposed method has been also tested on a set of images from the mini-Mammographic database provided by Mammographic Image Analysis Society (MIAS) publicly available.ResultsThe comparison between FCM-WF and standard FCM algorithms, applied on both databases, shows that the former produces better microcalcifications associations for clustering than the latter: with respect to the private and the public database we had a performance improvement of 10% and 5% with regard to the Merit Figure and a 22% and a 10% of reduction of false positives potentially identified in the images, both to the benefit of the FCM-WF. The method was also evaluated in terms of Sensitivity (93% and 82%), Accuracy (95% and 94%), FP/image (4% for both database) and Precision (62% and 65%).ConclusionsThanks to the private database and to the informations contained in it regarding every single microcalcification, we tested the developed clustering method with great accuracy. In particular we verified that 70% of the injected clusters of the private database remained unaffected if the reconstruction is performed with the FCM-WF. Testing the method on the MIAS databases allowed also to verify the segmentation properties of the algorithm, showing that 80% of pathological clusters remained unaffected.

Mammographic images segmentation based on chaotic map clustering algorithm

BackgroundThis work investigates the applicability of a novel clustering approach to the segmentation of mammographic digital images. The chaotic map clustering algorithm is used to group together similar subsets of image pixels resulting in a medically meaningful partition of the mammography.MethodsThe image is divided into pixels subsets characterized by a set of conveniently chosen features and each of the corresponding points in the feature space is associated to a map. A mutual coupling strength between the maps depending on the associated distance between feature space points is subsequently introduced. On the system of maps, the simulated evolution through chaotic dynamics leads to its natural partitioning, which corresponds to a particular segmentation scheme of the initial mammographic image.ResultsThe system provides a high recognition rate for small mass lesions (about 94% correctly segmented inside the breast) and the reproduction of the shape of regions with denser micro-calcifications in about 2/3 of the cases, while being less effective on identification of larger mass lesions.ConclusionsWe can summarize our analysis by asserting that due to the particularities of the mammographic images, the chaotic map clustering algorithm should not be used as the sole method of segmentation. It is rather the joint use of this method along with other segmentation techniques that could be successfully used for increasing the segmentation performance and for providing extra information for the subsequent analysis stages such as the classification of the segmented ROI.

Computer-Assisted Classification Patterns in Autoimmune Diagnostics: The AIDA Project

Antinuclear antibodies (ANAs) are significant biomarkers in the diagnosis of autoimmune diseases in humans, done by mean of Indirect ImmunoFluorescence (IIF) method, and performed by analyzing patterns and fluorescence intensity. This paper introduces the AIDA Project (autoimmunity: diagnosis assisted by computer) developed in the framework of an Italy-Tunisia cross-border cooperation and its preliminary results. A database of interpreted IIF images is being collected through the exchange of images and double reporting and a Gold Standard database, containing around 1000 double reported images, has been settled. The Gold Standard database is used for optimization of a CAD (Computer Aided Detection) solution and for the assessment of its added value, in order to be applied along with an Immunologist as a second Reader in detection of autoantibodies. This CAD system is able to identify on IIF images the fluorescence intensity and the fluorescence pattern. Preliminary results show that CAD, used as second Reader, appeared to perform better than Junior Immunologists and hence may significantly improve their efficacy; compared with two Junior Immunologists, the CAD system showed higher Intensity Accuracy (85,5% versus 66,0% and 66,0%), higher Patterns Accuracy (79,3% versus 48,0% and 66,2%), and higher Mean Class Accuracy (79,4% versus 56,7% and 64.2%).

Comparative Study of Human and Automated Screening for Antinuclear Antibodies by Immunofluorescence on HEp-2 Cells

Background : Several automated systems had been developed in order to reduce inter-observer variability in indirect immunofluorescence (IIF) interpretation. We aimed to evaluate the performance of a processing system in antinuclear antibodies (ANA) screening on HEp-2 cells. Patients and Methods : This study included 64 ANA-positive sera and 107 ANA-negative sera that underwent IIF on two commercial kits of HEp-2 cells (BioSystems® and Euroimmun®). IIF results were compared with a novel automated interpretation system, the “ Cyclopus CADImmuno®” (CAD). Results : All ANA-positive sera images were recognized as positive by CAD (sensitivity = 100%), while 17 (15.9%) of the ANA-negative sera images were interpreted as positive (specificity = 84.1%), κ=0.799 (SD=0.045). Comparison of IIF pattern determination between human and CAD system revealed on HEp-2 (BioSystems®), a complete concordance in 6 (9.37%) sera, a partial concordance (sharing of at least 1 pattern) in 42 (65.6%) cases and in 16 (25%) sera the pattern interpretation was discordant. Similarly, on HEp-2 (Euroimmun®) the concordance in pattern interpretation was total in 5 (7.8%) sera, partial in 39 (60.9%) and absent in 20 (31.25%). For both tested HEp-2 cells kits agreement was enhanced for the most common patterns, homogenous, fine speckled and coarse speckled. While there was an issue in identification of nucleolar, dots and nuclear membranous patterns by CAD. Conclusion : Assessment of ANA by IIF on HEp-2 cells using the automated interpretation system, the “ Cyclopus CADImmuno®” is a reliable method for positive/negative differentiation. Continuous integration of IIF images would improve the pattern identification by the CAD.

Automated approach for indirect immunofluorescence images classification based on unsupervised clustering method

Autoimmune diseases (ADs) are a collection of many complex disorders of unknown aetiology resulting in immune responses to self-antigens and are thought to result from interactions between genetic and environmental factors. ADs collectively are amongst the most prevalent diseases in the U.S., affecting at least 7% of the population. The diagnosis of ADs is very complex, the standard screening methods provides seeking and recognizing of Antinuclear Antibodies (ANA) by Indirect ImmunoFluorescence (IIF) based on HEp-2 cells. In this paper an automatic system able to identify and classify the Centromere pattern is presented. The method is based on the grouping of centromeres present on the cells through a clustering K-means algorithm. The performances were obtained on two public database of IIF images (A.I.D.A. and MIVIA). Our results showed a sensitivity for image of (90 ± 5)% and a Accuracy equal to (98.0 ± 0.5)%. Results demonstrate that the system is able to identify and classify Centromere pattern with accuracy better or comparable with some representative state of the art works. Moreover, it should be noted that for the classification phase the works used for the comparison used an expert-manual segmentation while, in the present work, the segmentation was obtained automatically.

Unsupervised clustering method for pattern recognition in IIF images

Autoimmune diseases are a family of more than 80 chronic, and often disabling, illnesses that develop when underlying defects in the immune system lead the body to attack its own organs, tissues, and cells. Diagnosis of autoimmune pathologies is based on research and identification of antinuclear antibodies (ANA) through indirect immunofluorescence (IIF) method and is performed by analyzing patterns and fluorescence intensity. We propose here a method to automatically classify the centromere pattern based on the grouping of centromeres on the cells through a clustering K-means algorithm. The described method was tested on a public database (MIVIA). The results of the test showed an Accuracy equal to (92.0 ± 1.0)%. Comparing our results with the results obtained on the MIVIA database it is possible to note that our method has a performance comparable with the three best values obtained. Indeed, the method here proposed allows an automatic segmentation and counting of the cells in the images, while the participants to the contest received the training set with the original images of the cells already segmented by specialists.