Table 1.

Information on two commercialized CAD systems

	AmCAD-UT	S-Detect for thyroid
		S-Detect 1	S-Detect 2
Technologies	Statistical pattern recognition and quantification algorithms	Support vector machine models using machine learning techniques	Convolutional neural network-based deep learning techniques
Characteristics	Application of CAD system by loading ultrasound images from PACS	Real-time application of CAD systems during US examinations
	Analysis of the sonographic characteristics (echogenic foci, echogenicity, texture, margin, anechoic areas, height/width ratio, nodule shape, and nodule size) and risk of malignancy based on the TI-RADS classifications	Analysis of the sonographic characteristics (composition, echogenicity, orientation, margin, spongiform, shape, calcifications, and nodule size) and presentation of a possible diagnosis using a dichotomous outcome (probably benign vs. probably malignant) or a TI-RADS classification outcome
	FDA 510(k) cleared	FDA approval in progress
Diagnostic performances	Similar sensitivity (87.0%), but lower specificity (68.8%) compared to those of clinical experts using the American Thyroid Association TI-RADS classification [29]	Comparable sensitivities (80.0%-92.0%), but lower specificity (74.6%-88.1%) compared to those of experienced radiologists using a dichotomous outcome [8,10-12,30]	Comparable sensitivities (81.4%), but lower specificity (68.2%-81.9%) compared to those of experienced radiologists [11,31]

CAD, computer-aided diagnosis; PACS, picture archiving and communication system; US, ultrasonography; TI-RADS, Thyroid Imaging Reporting and Data System; FDA, Food and Drug Administration.