The implementation of natural language processing to extract index lesions from breast magnetic resonance imaging reports

Table 1 Imaging features and their value set to be extracted from breast MRI reports

Entity type	Imaging features	Data type	Value set of imaging feature	Units
Mass	Location	Categorical	Right, Left, Upper, Outer quadrant …	N/A
	Shape	Categorical	Oval, Round, Irregular …	N/A
	Size	Numerical		mm
	Signal	Categorical		N/A
	T1WI	Categorical	Low signal, Isointensity, High signal	N/A
	T2WI	Categorical	Low signal, Isointensity, High signal	N/A
	DWI	Categorical	Low signal, Isointensity, High signal	N/A
	Margin	Categorical	Smooth, irregular, spiculated...	N/A
	Internal enhancement	Categorical	Homogeneous, Heterogeneous …	N/A
	Enhancement kinetic curve	Categorical	Persistent, Plateau, Wash-out	N/A
NME	Location	Categorical	Right, Left, Upper, Outer quadrant …	N/A
	Distribution pattern	Categorical	Focal area, Linear, Segmental …	N/A
	Scope	Numerical		mm
	Signal	Categorical		N/A
	T1WI	Categorical	Low signal, Isointensity, High signal	N/A
	T2WI	Categorical	Low signal, Isointensity, High signal	N/A
	DWI	Categorical	Low signal, Isointensity, High signal	N/A
	Internal enhancement	Categorical	Homogeneous, Heterogeneous...	N/A
	Enhancement kinetic curve	Categorical	Persistent, Plateau, Wash-out	N/A
Other associated findings	Lymphadenopathy	Categorical		N/A
	Invasion of skin, nipple, chest wall, pectoralis muscle	Categorical		N/A

T1WI T1-weighted imaging, T2WI T2-weighted imaging, DWI diffusion weighted imaging, NME non-mass enhancement, N/A not applicable

ISSN: 1472-6947