Multitask Cascade Convolution Neural Networks for Automatic Thyroid Nodule Detection and Recognition

Abstract

Thyroid ultrasonography is a widely-used clinical technique for nodule diagnosis in thyroid regions. However, it remains difficult to detect and recognize the nodules due to low contrast, high noise, and diverse appearance of nodules. In today's clinical practice, senior doctors could pinpoint nodules by analyzing global context features, local geometry structure, and intensity changes, which would require rich clinical experience accumulated from hundreds and thousands of nodule case studies. To alleviate doctors' tremendous labor in the diagnosis procedure, we advocate a machine learning approach to the detection and recognition tasks in this paper. In particular, we develop a multi-task cascade convolution neural network framework (MC-CNN) to exploit the context information of thyroid nodules. It may be noted that, our framework is built upon a large number of clinically-confirmed thyroid ultrasound images with accurate and detailed ground truth labels. Other key advantages of our framework result from a multi-task cascade architecture, two stages of carefully-designed deep convolution networks in order to detect and recognize thyroid nodules in a pyramidal fashion, and capturing various intrinsic features in a global-to-local way. Within our framework, the potential regions of interest after initial detection are further fed to the spatial pyramid augmented CNNs to embed multi-scale discriminative information for fine-grained thyroid recognition. Experimental results on clinical ultrasound images have indicated that, our MC-CNN is accurate and effective for both thyroid nodules detection and recognition. For the correct diagnosis rate of malignant and benign thyroid nodules, its mAP performance can achieve up to 98.2% accuracy, which outperforms the common CNNs by % on average. In addition, we conduct rigorous user studies to confirm that our MC-CNN outperforms experienced doctors, yet only consuming roughly 2% (1/48) of doctors' examination time on average. Therefore, the accuracy and efficiency of our new method exhibit its great potential in clinical applications.

Computer-Aided Diagnosis Systems in Diagnosing Malignant Thyroid Nodules on Ultrasonography: A Systematic Review and Meta-Analysis

“…The quality of the included studies is summarized in online supplementary Table 2. The risk of bias from patient selection was judged to be high or unclear in 13 of the included studies: 4 studies limited the nodule size within a certain scope [16,17,21,25]; 5 studies excluded difficult-to-diagnose nodules [15,[25][26][27]31]; and 4 studies were unclear about whether there were selected co-horts and inappropriate exclusions [14,19,23,29]. The risk of bias from the reference standard was considered to be unclear in 2 of the included studies [14,23].…”

“…The risk of bias from patient selection was judged to be high or unclear in 13 of the included studies: 4 studies limited the nodule size within a certain scope [16,17,21,25]; 5 studies excluded difficult-to-diagnose nodules [15,[25][26][27]31]; and 4 studies were unclear about whether there were selected co-horts and inappropriate exclusions [14,19,23,29]. The risk of bias from the reference standard was considered to be unclear in 2 of the included studies [14,23]. The risk of bias from flow and timing was considered to be high or unclear in 7 of the included studies, and these studies adopted pathological examination, fine needle aspiration, and ultrasound as reference standards for diagnosing benign nodules [19,22,25,27,28,30,31].…”

“…The flow diagram of the literature search is shown in Figure 1. Nineteen studies with 4,781 nodules used in external validation sets were included in the study, including 6 studies on classic machine learning-based CAD sys-DOI: 10.1159/000504390 tems [14][15][16][17][18][19] and 13 studies on deep learning-based CAD systems [7,[20][21][22][23][24][25][26][27][28][29][30][31]. The general characteristics of the included studies are shown in Table 1, and the detailed characteristics are demonstrated in online supplementary Table 1 (see www.karger.com/doi/10.1159/000504390 for all online suppl.…”

“…Thirteen studies with 1,667 malignant and 1,415 benign nodules were included in the analysis [7,[20][21][22][23][24][25][26][27][28][29][30][31]. The pooled sensitivity, specificity, AUC, and DOR are demonstrated in Figure 2b.…”

Deep Learning in Ultrasound Imaging

“…Similarly, in [106], [107], deep learning was used to accelerate echocardiographic exams by automatically recognizing the relevant standard views for further analysis, even permitting automated myocardial strain imaging [108]. In [109], a CNN was trained to perform thyroid nodule detection and recognition. Similar applications of deep learning include automated identification and segmentation of tumors in breast ultrasound [110], [111], [112], localization of clinically relevant B-line artifacts in lung ultrasonography [113], and real-time segmentation of anatomical zones on transrectal ultrasound (TRUS) scans [114].…”