Thyroid Nodule Recognition Based on Feature Selection and Pixel Classification Methods

Abstract

Statistical approach is a valuable way to describe texture primitives. The aim of this study is to design and implement a classifier framework to automatically identify the thyroid nodules from ultrasound images. Using rigorous mathematical foundations, this article focuses on developing a discriminative texture analysis method based on texture variations corresponding to four biological areas (normal thyroid, thyroid nodule, subcutaneous tissues, and trachea). Our research follows three steps: automatic extraction of the most discriminative first-order statistical texture features, building a classifier that automatically optimizes and selects the valuable features, and correlating significant texture parameters with the four biological areas of interest based on pixel classification and location characteristics. Twenty ultrasound images of normal thyroid and 20 that present thyroid nodules were used. The analysis involves both the whole thyroid ultrasound images and the region of interests (ROIs). The proposed system and the classification results are validated using the receiver operating characteristics which give a better overall view of the classification performance of methods. It is found that the proposed approach is capable of identifying thyroid nodules with a correct classification rate of 83 % when whole image is analyzed and with a percent of 91 % when the ROIs are analyzed.

Cascade convolutional neural networks for automatic detection of thyroid nodules in ultrasound images

“…6 These have led to extensive research in developing semi-automatically or automatically effective ultrasound image analysis techniques called computer aided detection (CAD) methods so as to obtain accurate, reproducible and more objective detection results. [7][8][9][10][11] The conventional CAD methods typically rely on the same three-step procedure: [12][13][14][15] (a) image pre-processing including denoising, enhancement, segmentation and so on; 16,17 (b) extracting and selecting the visual and effective features can represent major appearance components of the thyroid nodules; (c) detection by a classifier. Tsantis et al 7 utilized various morphological and wavelet-based features toward malignancy risk evaluation of thyroid nodules based on support vector machines (SVMs) and probabilistic neural networks (PNNs).…”

Multi‐institutional validation of a novel textural analysis tool for preoperative stratification of suspected thyroid tumors on diffusion‐weighted MRI

“…Several TA studies show good discrimination of thyroid nodules on ultrasound images (13–15) and better distinction between benign and malignant thyroid lesions on nuclear chromatin images (16), but none have utilized TA on DW-MRI scans of the thyroid. The aim of this study was to assess whether textural analysis could improve the accuracy, sensitivity, and specificity of DW-MRI for the stratification of malignancy in suspicious thyroid nodules.…”

“…Textural analysis (TA) has become an attractive clinical tool, as it quantifies pixel intensity variation otherwise invisible to the naked eye and thus aids in characterizing underlying tissue structures. Several TA studies have shown good discrimination of thyroid nodules on ultrasound images (13)(14)(15) and better distinction between benign and malignant thyroid lesions on nuclear chromatin images (16), but none have used TA on DW-MRI scans of the thyroid. The aim of this study was to assess whether textural analysis could improve the accuracy, sensitivity, and specificity of DW-MRI for the stratification of malignancy in suspicious thyroid nodules.…”

Quantitative multivariate analysis of dynamic multicellular morphogenic trajectories

“…Our method reconstructs cellular locations as interacting networks that can subsequently be further subdivided into biologically relevant sub-networks. This network-based approach circumvents problems associated with traditional classification methods that rely solely on standardized images 19 and use of individual pixel classification methodologies 36, 37 . While some systems exist for classifying spatial patterns in zebrafish 17 , C. elegans 15 and Drosophila embryos 19,38–40 , previous approaches require specifically orientated and annotated images, are specific to the organism of interest, and/or often do not have single cell resolution.…”