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 multitask cascade convolution neural network (MC-CNN) framework 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 multitask 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 multiscale discriminative information for fine-grained thyroid recognition. Experimental results on 4309 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 mean Average Precision (mAP) performance can achieve up to [Formula: see text] accuracy, which outperforms the common CNNs by [Formula: see text] on average. In addition, we conduct rigorous user studies to confirm that our MC-CNN outperforms experienced doctors, yet only consuming roughly [Formula: see text] ( 1/48) of doctors' examination time on average. Therefore, the accuracy and efficiency of our new method exhibit its great potential in clinical applications.

An Analog Bayesian Classifier Implementation, for Thyroid Disease Detection, based on a Low-Power, Current-Mode Gaussian Function Circuit

Low-Power Analog Integrated Architecture of the Voting Classification Algorithm for Diabetes Disease Prediction

A Low-Power Analog Integrated Euclidean Distance Radial Basis Function Classifier

Nanopower Integrated Gaussian Mixture Model Classifier for Epileptic Seizure Prediction

A Two-Stage Detection Methodology for Thyroid Cancer Using Photonic Crystal: Logistic Regression and Artificial Neural Networks

Real-time ultra-low power ECG anomaly detection using an event-driven neuromorphic processor

An Energy-Efficient ECG Processor with Weak-Strong Hybrid Classifier for Arrhythmia Detection

Thyroid Disease Classification using Machine Learning Algorithms

Low-power perceptron model based ECG processor for premature ventricular contraction detection.

Multi-channel convolutional neural network architectures for thyroid cancer detection

Artificial Intelligence-Based Thyroid Nodule Classification Using Information from Spatial and Frequency Domains

A Low-Power Analog Integrated Implementation of the Support Vector Machine Algorithm with On-Chip Learning Tested on a Bearing Fault Application

A Low-Power ECG Processor ASIC Based on an Artificial Neural Network for Arrhythmia Detection

Multiple-Instance Learning for thyroid gland disease classification: A hands-on experience

A Hand Gesture Recognition Circuit Utilizing an Analog Voting Classifier

A novel maternal thyroid disease prediction using multi-scale vision transformer architecture with improved linguistic hedges neural-fuzzy classifier

EKGNet: A 10.96μW Fully Analog Neural Network for Intra-Patient Arrhythmia Classification

A Flexible Low-Power Machine Learning Accelerator for Healthcare Applications

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

A 746 Nw ECG Processor ASIC Based on Ternary Neural Network

Empirical Method for Thyroid Disease Classification Using a Machine Learning Approach