Discriminant analysis and machine learning approach for evaluating and improving the performance of immunohistochemical algorithms for COO classification of DLBCL

Table 1 Performance metrics of classification of IHC-decision tree algorithms and LDA

	Algorithm	Antibody combination	Acc	Sens	Spec	PPV	NPV	LR+	LR−
IHC-decision trees	Nyman	3,5	0.72	0.52	0.91	0.84	0.67	5.56	0.53
	Colomo	1,2,5	0.78	0.71	0.84	0.81	0.75	4.56	0.34
	Hans	1,2,5	0.85	0.91	0.78	0.80	0.91	4.21	0.11
	Hans*	1,5	0.82	0.94	0.70	0.75	0.92	3.14	0.09
	Choi	1,2,3,4,5	0.88	0.94	0.84	0.84	0.93	5.70	0.08
	Choi*	1,3,4,5	0.79	0.74	0.83	0.80	0.77	4.30	0.31
	VY3	1,2,3	0.88	0.92	0.84	0.85	0.92	5.92	0.09
	VY4	1,2,3,4	0.88	0.93	0.84	0.85	0.92	5.80	0.09
Linear discriminant analysis	As in Hans*	1,5	0.84	0.77	0.91	0.89	0.81	8.59	0.25
	As in Nyman	3,5	0.77	0.81	0.74	0.75	0.81	3.10	0.25
	As in VY3	1,2,3	0.89	0.87	0.91	0.90	0.88	9.19	0.15
	As in Hans/Colomo	1,2,5	0.87	0.86	0.88	0.87	0.87	7.25	0.16
	–	1,4,5	0.87	0.81	0.92	0.90	0.84	9.93	0.20
	As in VY4	1,2,3,4	0.87	0.84	0.90	0.89	0.86	8.24	0.17
	As in Choi*	1,3,4,5	0.88	0.86	0.91	0.90	0.87	9.09	0.16
	As in Choi	1,2,3,4,5	0.89	0.87	0.91	0.90	0.88	9.23	0.14

The upper section corresponds to the performance of the IHC-decision tree algorithms. Lower section corresponds to equivalent combinations of antibodies, but with LDA classification, this includes the rest of combinations not reported by IHC-decision tree algorithms. Choi, VY3, and VY4 algorithms reached the most considerable accuracy, representing the most balanced options of sensibility and specificity, with similar performance metrics
Numeric tags 1 = CD10, 2 = BCL6, 3 = FOXP1, 4 = GCTE1, and 5 = MUM1
Acc: accuracy; Sens: sensitivity; Spec: specificity; PPV: positive predictive value; NPV: negative predictive values; LR+: likelihood ratio for positive test results; LR−: likelihood ratio for negative test result

ISSN: 1479-5876