Comparison of Artificial Neural Network with Logistic Regression as Classification Models for Variable Selection for Prediction of Breast Cancer Patient Outcomes

Table 1

Variables included in the database.

Variables	Modalities
Breast side	Right, Left
Age	Real
Delay diagnosis—1st treatment	Time
Personal cancer history	No, Yes
Menopausal status	No, Yes
Cutaneous inflammation	No, Yes
Skin invasion	No, Yes
Clinical number of nodules	0, 1, ≥2
Clinical T stage (size)	T0, T1, T2, T3, T4
Clinical N stage	N0, N+
Number of tumours (histology)	0 or 1, ≥2
Tumour size (mm) (histology)	Real
Histological type	Lobular, Ductal, Mixed, Microinvasive, Others
SBR grade	1, 2, 3
Lymphatic embolus	No, Yes
Nervous colonisation	No, Yes
Necrosis (histology)	No in situ component,
	In situ component without necrosis
	In situ component with necrosis
Infiltrating tumour associated with CIS	Others, >75% in situ
Limits of exeresis—infiltrating carcinoma	Not in sano, In sano
Skin—infiltrating tumour	No or Not Applicable, Yes
Skin—embolus	No or Not Applicable, Yes
Nipple—infiltrating tumour	No or Not Applicable, Yes
Nipple—cancer in situ	No or Not Applicable, Yes
Nipple—embolus	No or Not Applicable, Yes
Nipple—Paget’s disease	No or Not Applicable, Yes
Pectoral muscle invaded	No or Not Applicable, Yes
Number of invaded nodes	0, 1 or 2, ≥3
Nodes invaded	N0, Micro metastasis, Macrometastasis
Capsular breaking	No, Yes
Oestrogen receptor: % marked cells (count)	<10%, (10–50)%, >50%
Oestrogen receptor (intensity)	+, ++/+++
Progesterone receptor: % marked cells (count)	<10%, [10-50]%, >50%
Progesterone receptor (intensity)	+, ++/+++