Machine learning prediction of pathological complete response and overall survival of breast cancer patients in an underserved inner-city population

Table 8 Multivariate results of top predictors of OS for all 4 models utilizing top 10 predictors including MRI data (N = 240)

	Predictors	Acc	Spec	Sens	AUC
Neural network	Tumor size, T-stage, nipple involvement, N stage, pCR, triple negative, skin involvement, lymph node involvement, ER+/HER2, pectoralis muscle involvement	0.947 ± 0.022	0.986 ± 0.019	0.062 ± 0.076	0.840 ± 0.117
Random forest	Tumor size, T-stage, N stage, pCR, triple negative, age, nipple involvement, skin involvement, lymph node involvement, ER+/HER2	0.915 ± 0.035	0.932 ± 0.025	0.667 ± 0.209	0.830 ± 0.045
Logistic regression	Tumor size, T-stage, pCR, N stage, triple negative, nipple involvement, skin involvement, age, ER+/HER2, pectoralis muscle involvement	0.898 ± 0.055	0.985 ± 0.012	0.310 ± 0.284	0.850 ± 0.098
Gradient boosted regression	Tumor size, T-stage, pCR, triple negative, nipple involvement, N stage, skin involvement, lymph node involvement, age, ER+/HER2	0.869 ± 0.306	0.953 ± 0.031	0.244 ± 0.167	0.841 ± 0.118

ISSN: 1465-542X