Parametrical modelling for texture characterization—A novel approach applied to ultrasound thyroid segmentation

Abstract

Texture analysis is an important topic in Ultrasound (US) image analysis for structure segmentation and tissue classification. In this work a novel approach for US image texture feature extraction is presented. It is mainly based on parametrical modelling of a signal version of the US image in order to process it as data resulting from a dynamical process. Because of the predictive characteristics of such a model representation, good estimations of texture features can be obtained with less data than generally used methods require, allowing higher robustness to low Signal-to-Noise ratio and a more localized US image analysis. The usability of the proposed approach was demonstrated by extracting texture features for segmenting the thyroid in US images. The obtained results showed that features corresponding to energy ratios between different modelled texture frequency bands allowed to clearly distinguish between thyroid and non-thyroid texture. A simple k-means clustering algorithm has been used for separating US image patches as belonging to thyroid or not. Segmentation of thyroid was performed in two different datasets obtaining Dice coefficients over 85%.

Operational framework and training standard requirements for AI‐empowered robotic surgery

“…We, therefore, propose that one can add audio sensors to the robotic systems that measure the signal between a tool‐tissue interaction, and other similar existing approaches should be taken into consideration 21 . Advanced signal processing could extract features that can be used to improve device guiding accuracy and may even be able to classify tissues 22,23 . The obtained cadaver audio profiles of different tissues could be used as base information to create a virtual tissue histology database.…”

Thyroid Ultrasound Texture Classification Using Autoregressive Features in Conjunction With Machine Learning Approaches

“…Each image from the two datasets were first divided into non-overlapping texture patches of size 20 x 20 and following the ground truth, each patch was labelled either thyroid (=0) or non-thyroid (=1). The size of the texture patch was set in such a way that it captured important dynamical changes that allowed to involve a number of main frequency components that can help to spectrally differentiate thyroid and non-thyroid regions (see [3]). On top of that, it should also cover all the smaller regions inside the thyroid (for example the isthmus region as marked by yellow solid lines in Fig.…”

“…In this section, we will mainly discuss on how the features were computed from the thyroid images which were used for the training of the classifiers for thyroid texture classification. A detailed explanation on AR modelling, feature computation and prominent features selection have been explained in our recent work [3] but we will only introduce the main steps here. We used AR modelling to compute the features from the texture patches.…”

“…A set of 30 different features were computed from the AR parameters using the energy ratios between different frequency bands for each texture patch in the texture database. A detailed explanation on AR modelling, features computation and prominent features selection have been explained in [3]. A flowchart representing the entire feature computation process is shown in Fig.…”

“…The features computed in this work for training the classifier are based on a novel texture characterization algorithm published previously by our team [3]. A signal based parametrical approach using Autoregressive (AR) modelling has been proposed to characterize the thyroid texture using 30 AR spectral energy ratios based features that can distinguish between thyroid and non-thyroid regions.…”