Retrospective validation study of an artificial neural network-based preoperative decision-support tool for noninvasive lymph node staging (NILS) in women with primary breast cancer (ISRCTN14341750)

Table 3 Comparison between performance measures of the noninvasive lymph node staging (NILS) model, including potential sentinel lymph node biopsy (SLNB) reduction rates between the original ANN model and current model

Model	TP	TN	FP	FN	Sensitivity	Specificity	FNR^a	SLNB reduction rate^b
ANN Prototype (2019)	258	190	324	28	91%	37%	10%	27%
NILS (1.0)	256	177	339	28	90%	34%	10%	26%
NILS (version 0.1.0) Validation	136	133	299	18	88%	31%	12%	26%

^aEquivalent to the maximum FNR of 10% reflecting accepted FNR of the SLNB procedure
^bThe SLNB reduction rate was calculated as follows = (TN + FN)/(TN + FN + TP + FP)
Abbreviations: TP True positive, TN True negative, FP False positive, FN False negative, FNR False negative rate, SLNB Sentinel lymph node biopsy

ISSN: 1471-2407