Carolina M. Azevedo

Complexity curve and grey level co‐occurrence matrix in the texture evaluation of breast tumor on ultrasound images

This work aims at investigating texture parameters in distinguishing malign and benign breast tumors on ultrasound images. A rectangular region of interest (ROI) containing the tumor and its neighboring was defined for each image. Five parameters were extracted from the complexity curve (CC) of the ROI. Another five parameters were calculated from the grey-level co-occurrence matrix (GLCM) also for the ROI. The same was carried out for internal tumor region, hence, totaling 20 parameters. The linear discriminant analysis was applied to sets of up to five parameters and then the performances were assessed. The most relevant individual parameters were the contrast (con) (from the GLCM over the ROI) and the maximum value (mvi) from the CC just for the tumor internal region). When they were taken together, a correct classification slightly over 80% of the breast tumors was achieved. The highest performance (accuracy=84.2%, sensitivity=87.0%, and specificity=78.8%) was obtained with mvi, con, the standard deviation of the pixel pairs and the entropy, both for GLCM, and the internal region contrast also from GLCM. Parameters extracted from the internal region generally performed better and were more significant than those from the ROI. Moreover, parameters calculated only from CC or GLCM resulted in no statistically significant performance difference. These findings suggest that the texture parameters can be useful to help radiologist in distinguishing between benign or malign breast tumors on ultrasound images.

Assessing the performance of morphological parameters in distinguishing breast tumors on ultrasound images

This work aims at investigating seven morphological parameters in distinguishing malignant and benign breast tumors on ultrasound images. Linear discriminant analysis was applied to sets of up to five parameters and then the performances were assessed using the area Az (+/- standard error) under the ROC curve, accuracy (Ac), sensitivity (Se), specificity (Sp), positive predictive value and negative predictive value. The most relevant individual parameters were the normalized residual value (nrv) and overlap ratio (RS), both calculated from the convex polygon technique, and the circularity (C). When nrv and C were taken together with roughness (R), calculated from normalized radial length (NRL), a performance slightly over 83% in distinguishing malignant and benign breast tumors was achieved.

Assessing the combined performance of texture and morphological parameters in distinguishing breast tumors in ultrasound images

PURPOSE This work aims to investigate the combination of morphological and texture parameters in distinguishing between malignant and benign breast tumors in ultrasound images. METHODS Linear discriminant analysis was applied to sets of up to five parameters, and then the performances were assessed using the area A(z) (± standard error) under the receiver operator characteristic curve, accuracy (Ac), sensitivity (Se), specificity (Sp), positive predictive value, and negative predictive value. RESULTS The most relevant individual parameter was the normalized residual value (nrv), calculated from the convex polygon technique. The best performance among all studied combinations was achieved by two morphological and three texture parameters (nrv, con, std, R, and asm(i)), which correctly distinguished nearly 85% of the breast tumors. CONCLUSIONS This result indicates that the combination of morphological and texture parameters may be useful to assist physicians in the diagnostic process, especially if it is associated with an automatic classification tool.

Classifying Breast Tumours on Ultrasound Images Using a Hybrid Classifier and Texture Features

This work aims to classify breast tumours on ultrasound (US) images, using texture features calculated from complexity curve (CC) and grey-level co-occurence matrix (GLCM), applied to a proposed hybrid classifier based on a multilayer perceptron (MLP) network and genetic algorithms (GA). A rectangular region of interest (ROI) containing the tumour and its neighbouring is defined for each image. Five features are extracted from CC of the ROI, and another five are calculated from GLCM also for the ROI. The same is obtained for internal tumour region, hence totalling 20 parameters. The hybrid classifier uses GA to select the best set of input features, limited up to 5, while MLP is trained by the backpropagation algorithm. The leave- one-case-out re-sampling method is carried out to assure the reliability and effectiveness of the classifier. The results are compared to the ones presented in a previous work, where Fishers Linear Discriminant Analysis (LDA) was applied. The proposed hybrid classifier achieved a global performance superior to 90.0% and statistically significant higher than LDA. Hence, our findings suggest that the combination of texture features and the hybrid classifier can aid radiologists in making the diagnoses of malignant breast tumours on US images.

Análise computacional da textura de tumores de mama em imagens por ultrassom de pacientes submetidas a cirurgia conservadora

OBJECTIVE: The purpose of this study was to assess the features of breast lesion texture on sonographic images of patients submitted to breast-conserving surgery, with or without tumor recurrence. MATERIALS AND METHODS: Sonographic images of 36 patients submitted to conservative surgery for breast cancer, 12 of them with, and 24 without local recurrence, included 3 contralateral malignant lesions, 7 benign lumps (3 cysts and 4 fibroadenomas), 5 atypical hyperplasias and 9 fibrocystic changes. The quantification of features of breast lesion texture was based on ten parameters calculated from gray-level co-occurrence matrix and complexity curve. Linear discriminant analysis was applied to the texture parameters for differentiating between breast lesions in women submitted to conservative surgery with and without tumor recurrence. RESULTS: The assessment of performance of texture parameters in distinguishing lesion recurrences in a group including benign lumps and atypical hyperplasias demonstrated specificity of 100%, and accuracy and sensitivity > 91%. Another test demonstrated that texture parameters were useful in the differentiation between atypical hyperplasias and benign lumps. CONCLUSION: Despite the limited number of cases, the present results can be considered as promising, suggesting that texture parameters may help in the differentiation among benign lumps, atypical hyperplasias and recurrent malignant lesions.

Aspectos tomográficos da órbita aguda infecciosa: revisão de literatura

The acute and nontraumatic diseases that involve the orbit are often little known by most physicians. These conditions are due to several factors, such as immune disorders, congenital, infections, vascular, among others disorders. The infectious causes correspond to more than 50% of all cases and require rapid diagnosis and management in order to minimize sequels. Computed tomography (CT) is the first line imaging method on these cases, generally being available in emergency centers and capable to provide an accurate, quick and effective diagnostic information. This review article aims to describe the main tomographic findings in acute orbit infections, correlating them with the literature data.

Morphometric and Texture Parameters in Distinguishing Breast Tumours on Ultrasound Images

Ultrasound (US) images associated with mammography help radiologists in establishing the diagnostic of breast cancer. Therefore, the extraction of morphometric (MP) and texture (TP) parameters from breast tumour US images has been largely investigated. But the debate about best parameters choice still persists. Literature reports limitations on TP calculation, as its sensibility to nonlinear variations with US system setting and focal depth. MPs also have limitations but regarding tumour contour extraction process. Nevertheless, relevant results can be achieved using TPs, fixing or not the US system setting, and using MPs calculated from manual or automatic contour estimation. This work investigates performance of seven MP and 20 TP for distinguishing malignant breast tumour US images. With this aim, a segmentation based on Mathematical Morphology was applied to 152 breast US images. Linear Discriminant Analysis (LDA) was carried out with all MP and TP combinations. Furthermore, the area under ROC curve (Az) was used to assess the performance. The normalised residual mean square value (introduced in this work) combined with two other MPs, resulted the highest Az (0.92), accuracy (Ac=88.8%), sensitivity (Se=88.0%) and specificity (Sp=90.4%). With 12 TPs (calculated from co-occurrence matrix and complexity curve), the performance was Az=0.90, Ac=88.2%, Se=89.0% and Sp=86.5%. Considering MP and TP combined, the highest discrimination performance (Az=0.93, Ac=92.8%, Se=92.0% and Sp=94.2%) was obtained when 5 MPs and 6 TPs were taking together. This finding suggests that the combination of MP and TP of US images can be helpful in breast tumours diagnosis.

GA-BACKPROPAGATION HYBRID TRAINING AND MORPHOMETRIC PARAMETERS TO CLASSIFY BREAST TUMOURS ON ULTRASOUND IMAGES

Abstract This work presents a multilayer perceptron (MLP) network, trained with backpropagation algorithm, to classify breast tumours as malign or benign ones. Seven morphometric parameters, extracted from the convex polygon and the normalised radial length techniques, are used as MLP input. A genetic-based selection procedure helps backpropagation training scheme to select the best input parameters and best training set, as well. To achieve this aim, an objective function is proposed. The best values of accuracy (97.4%), sensitivity (98.0%) and specificity (96.2%) were achieved with a set of five parameters, despite the training set sizes tested: 30% and 50% of the total samples.