Table 1 Performance summary of the SDAE, CURVE and RANK algorithms on the lung CT dataset.

From: Computer-Aided Diagnosis with Deep Learning Architecture: Applications to Breast Lesions in US Images and Pulmonary Nodules in CT Scans

LUNG		ACC (%)	AUC (%)	SENS (%)	SPEC (%)	PPV (%)	NPV (%)
SDAE1 SINGLE	cv-all	87.4 ± 3.3	94.1 ± 1.9	86.3 ± 5.6	88.5 ± 4.8	86.9 ± 4.5	88.8 ± 4.1
	cv1	88.6 ± 2.5	95.0 ± 2.1	88.1 ± 3.8	88.0 ± 5.5	89.2 ± 3.1	88.4 ± 4.3
	cv5	87.3 ± 3.2	93.5 ± 1.9	86.9 ± 4.3	87.7 ± 5.8	87.1 ± 3.5	89.4 ± 3.6
	cv10	86.6 ± 4.1	93.5 ± 1.7	87.4 ± 4.2	85.9 ± 7.4	87.3 ± 3.8	88.1 ± 4.8
CURVE1 SINGLE	cv-all	77.8 ± 2.9	86.0 ± 2.6	75.9 ± 4.9	79.6 ± 3.6	76.9 ± 3.8	78.9 ± 3.1
	cv1	78.4 ± 2.3	86.1 ± 2.9	76.6 ± 4.4	80.1 ± 3.5	77.5 ± 3.2	79.5 ± 2.8
	cv5	77.4 ± 4.6	86.0 ± 4.0	76.1 ± 8.2	78.6 ± 4.1	77.1 ± 6.6	78.0 ± 3.6
	cv10	77.2 ± 3.1	85.7 ± 2.9	75.9 ± 5.1	78.6 ± 4.4	76.6 ± 3.9	78.1 ± 3.7
RANK1 SINGLE	cv-all	75.9 ± 3.1	81.8 ± 3.4	73.8 ± 4.9	78.0 ± 4.4	75.0 ± 3.6	77.1 ± 3.6
	cv1	75.6 ± 3.5	81.5 ± 3.5	73.9 ± 5.8	77.4 ± 4.6	74.9 ± 4.3	76.7 ± 3.8
	cv5	75.5 ± 1.6	81.6 ± 3.7	73.1 ± 3.8	77.9 ± 3.5	74.5±2.4	76.9 ± 2.5
	Cv10	76.0 ± 3.9	82.0 ± 4.7	74.0 ± 5.3	78.0 ± 4.8	75.1 ± 4.3	77.2 ± 4.3
SDAE2 ALL	cv-all	94.4 ± 3.2	98.4 ± 1.5	90.8 ± 5.3	98.1 ± 2.2	91.6 ± 4.4	97.9 ± 2.5
	cv1	95.6 ± 3.0	98.9 ± 1.0	92.4 ± 5.4	98.9 ± 1.3	93.1 ± 4.6	98.8 ± 1.4
	cv5	94.6 ± 2.5	98.5 ± 1.7	90.9 ± 4.1	98.3 ± 2.3	91.6 ± 3.5	98.3 ± 2.4
	cv10	93.6 ± 3.1	98.4 ± 1.1	89.9 ± 4.2	97.3 ± 2.7	90.6 ± 3.7	97.3 ± 2.9
CURVE2 ALL	cv-all	78.4 ± 2.9	85.1 ± 2.7	75.1 ± 4.5	81.7 ± 4.2	76.7 ± 3.3	80.5 ± 3.7
	cv1	78.9 ± 3.1	85.0 ± 2.5	75.3 ± 6.1	82.4 ± 2.5	77.2 ± 4.3	81.1 ± 2.5
	cv5	78.4 ± 3.8	85.2 ± 3.1	74.3 ± 4.8	82.6 ± 5.3	76.3 ± 3.7	81.2 ± 5.0
	cv10	77.1 ± 1.5	84.7 ± 3.1	74.1 ± 4.1	80.1 ± 4.2	75.7 ± 2.1	79.1 ± 3.0
RANK2 ALL	cv-all	72.0 ± 3.5	74.6 ± 4.1	62.0 ± 5.4	82.0 ± 4.2	68.4 ± 3.3	77.6 ± 4.4
	cv1	72.5 ± 3.6	74.9 ± 5.0	63.3 ± 5.5	81.7 ± 4.2	69.1 ± 3.4	77.7 ± 4.6
	cv5	71.8 ± 2.9	74.4 ± 3.5	61.7 ± 4.4	81.9 ± 5.4	68.2 ± 2.4	77.6 ± 4.9
	Cv10	72.0 ± 2.5	74.1 ± 2.7	61.4 ± 3.6	82.6 ± 2.8	68.2 ± 2.3	77.9 ± 3.2
MORPH MAN	cv-all	78.1 ± 3.3	86.6 ± 3.0	70.9 ± 5.1	88.5 ± 4.8	74.7 ± 3.4	83.1 ± 4.3
	cv1	78.1 ± 4.1	86.5 ± 3.0	70.9 ± 4.2	88.0 ± 5.5	74.6 ± 3.5	83.1 ± 5.8
	cv5	78.4 ± 3.4	86.7 ± 2.8	71.0 ± 6.9	87.7 ± 5.8	75.0 ± 4.2	83.4 ± 2.4
	Cv10	78.0 ± 1.8	86.6 ± 2.3	70.7 ± 3.6	85.9 ± 7.4	74.5 ± 1.9	83.0 ± 3.8
MORPH DRLSE	cv-all	71.8 ± 3.7	76.2 ± 3.6	63.3 ± 5.7	80.4 ± 4.4	68.8 ± 3.6	76.5 ± 4.5
	cv1	72.1 ± 2.8	76.3 ± 2.5	63.6 ± 5.7	80.7 ± 3.8	69.0 ± 3.3	76.8 ± 3.3
	cv5	71.5 ± 5.0	76.0 ± 5.1	62.7 ± 8.4	80.3 ± 4.6	68.5 ± 5.3	76.0 ± 4.8
	Cv10	71.8 ± 3.7	76.0 ± 3.4	63.4 ± 5.0	80.1 ± 4.4	68.7 ± 3.5	76.1 ± 4.1
MORPH GC	cv-all	73.3 ± 3.7	78.8 ± 3.3	66.2 ± 5.4	80.3 ± 4.8	70.5 ± 3.6	77.2 ± 4.5
	cv1	73.5 ± 1.8	78.7 ± 1.7	66.7 ± 3.3	80.3 ± 3.6	70.7 ± 1.8	77.3 ± 2.8
	cv5	73.6 ± 3.5	78.9 ± 3.6	66.4 ± 7.0	80.7±3.6	70.8 ± 4.0	77.5 ± 3.3
	Cv10	73.1 ± 3.7	78.7 ± 3.3	66.0 ± 3.3	80.3 ± 3.6	70.3 ± 2.5	77.1 ± 3.7
CURVE1 DRLSE SINGLE	cv-all	75.7 ± 3.7	83.9 ± 3.5	67.3 ± 6.3	84.0 ± 5.0	72.2 ± 3.9	81.0 ± 4.8
	cv1	76.4 ± 3.8	85.3 ± 4.1	69.0 ± 5.3	83.9 ± 5.0	73.1 ± 3.7	81.2 ± 5.2
	cv5	75.8 ± 2.7	83.1 ± 3.2	67.3 ± 3.0	84.3 ± 3.0	72.0 ± 2.4	81.1 ± 3.4
	Cv10	74.3 ± 3.5	83.2 ± 2.9	65.9 ± 8.7	82.7 ± 3.9	71.2 ± 5.0	79.3 ± 3.1
CURVE1 GC SINGLE	cv-all	76.3 ± 3.7	84.4 ± 3.4	67.0 ± 6.3	85.6 ± 4.8	72.3 ± 3.9	82.5 ± 4.8
	cv1	78.0 ± 3.9	85.0 ± 4.0	69.7 ± 3.6	86.3 ± 5.4	74.0 ± 3.1	83.7 ± 5.8
	cv5	75.9 ± 2.7	83.9±3.0	66.3 ± 5.3	85.8 ± 3.1	71.9 ± 3.3	82.2 ± 3.1
	Cv10	75.4 ± 1.7	84.3 ± 2.4	65.6 ± 5.6	85.3 ± 3.6	71.4 ± 2.7	81.9 ± 2.8
RANK1 DRLSE SINGLE	cv-all	75.9 ± 4.2	82.7 ± 4.2	74.9 ± 5.9	76.9 ± 6.7	75.5 ± 4.5	76.7 ± 5.3
	cv1	77.5 ± 5.2	84.0 ± 4.9	76.1 ± 5.6	78.9 ± 9.0	76.8 ± 4.7	78.9 ± 9.0
	cv5	74.5 ± 3.2	81.8 ± 3.9	75.6 ± 3.8	73.4 ± 3.8	75.1 ± 3.5	73.4 ± 3.8
	Cv10	76.5 ± 3.8	82.5 ± 2.7	75.7 ± 4.8	77.3 ± 5.5	76.2 ± 4.0	77.3 ± 5.5
RANK1 GC SINGLE	cv-all	75.4 ± 4.0	82.0 ± 4.0	72.1 ± 6.0	78.7 ± 5.7	74.0 ± 4.3	77.4 ± 4.8
	cv1	76.2 ± 4.7	83.0 ± 5.2	71.6 ± 7.2	80.9 ± 6.2	74.2 ± 5.0	79.2 ± 6.1
	cv5	75.3 ± 4.8	81.5 ± 3.0	72.9 ± 5.1	77.7 ± 7.0	74.1 ± 4.2	76.8 ± 5.7
	Cv10	75.1 ± 4.1	81.6 ± 3.2	72.7 ± 5.9	77.6 ± 5.8	74.1 ± 4.3	76.6 ± 4.9
CURVE2 DRLSE ALL	cv-all	76.6 ± 3.1	85.4 ± 2.8	68.0 ± 5.5	85.2 ± 4.2	72.8 ± 3.4	82.3 ± 4.1
	cv1	77.2 ± 2.7	85.3 ± 2.6	68.6 ± 5.8	85.9 ± 2.6	73.4 ± 3.4	82.9 ± 2.5
	cv5	76.3 ± 3.9	85.3 ± 2.3	68.0 ± 5.5	84.6 ± 3.4	72.6 ± 3.9	81.5 ± 4.0
	Cv10	76.3 ± 3.8	85.5 ± 3.6	68.0 ± 5.6	84.6 ± 5.5	72.7 ± 3.8	81.8 ± 5.2
CURVE2	cv-all	78.3 ± 3.5	85.5 ± 3.0	75.7 ± 5.5	81.0 ± 4.2	77.1 ± 4.2	80.0 ± 3.8
GC	cv1	78.6 ± 4.6	85.4 ± 3.8	76.1 ± 4.5	81.1 ± 5.8	77.3 ± 4.2	80.3 ± 5.6
ALL	cv5	78.6 ± 4.0	85.3 ± 3.2	76.3 ± 5.9	80.9 ± 3.6	77.5 ± 4.6	79.9 ± 3.7
	Cv10	78.2 ± 4.0	85.2 ± 2.4	75.1 ± 6.5	81.3 ± 2.6	76.8 ± 4.9	80.0 ± 3.1
RANK2 DRLSE ALL	cv-all	69.1 ± 3.7	83.1 ± 3.3	48.2 ± 6.4	90.0 ± 3.8	63.6 ± 3.1	82.9 ± 5.6
	cv1	69.6 ± 3.3	82.9 ± 2.6	48.7 ± 5.4	90.4 ± 3.6	63.9 ± 2.6	83.7 ± 5.4
	cv5	69.1 ± 2.8	83.6 ± 2.9	48.1 ± 5.4	90.1 ± 2.8	63.5 ± ± 2.3	83.1 ± 4.1
	Cv10	69.1 ± 4.2	83.2 ± 3.1	48.1 ± 7.5	90.0 ± 3.9	63.6 ± 3.7	82.9 ± 5.6
RANK2 GC ALL	cv-all	72.2 ± 3.3	83.8 ± 3.1	57.8 ± 5.8	86.5 ± 4.1	67.3 ± 3.0	81.3 ± 4.6
	cv1	73.0 ± 2.2	83.9 ± 2.4	58.6 ± 4.2	87.4 ± 3.3	67.9 ± 2.2	82.5 ± 3.6
	cv5	72.5 ± 2.6	83.8 ± 1.7	58.1±5.7	86.9 ± 3.4	67.6 ± 2.7	81.7 ± 3.3
	Cv10	71.6 ± 2.6	84.1 ± 3.1	57.3 ± 6.2	86.0 ± 2.5	67.0 ± 2.8	80.4 ± 2.4
CURVE1 MAN SINGLE	cv-all	77.0 ± 3.6	85.9 ± 2.9	69.3 ± 5.8	84.7 ± 4.0	73.5 ± 3.9	82.0 ± 4.2
	cv1	76.6 ± 4.8	85.6 ± 3.4	69.3 ± 6.5	84.3 ± 3.0	73.4 ± 4.8	81.4 ± 5.7
	cv5	76.6 ± 3.7	85.6 ± 2.7	68.9 ± 6.2	84.0 ± 5.4	73.2 ± 3.0	81.4 ± 3.4
	Cv10	76.4 ± 3.2	85.6 ± 2.6	68.1 ± 3.8	84.7 ± 3.6	72.7 ± 2.9	81.7 ± 4.1
RANK1 MAN SINGLE	cv-all	74.9 ± 3.3	83.9 ± 3.0	66.6 ± 5.5	83.2 ± 4.5	71.5 ± 3.4	80.0 ± 4.2
	cv1	75.7 ± 3.4	83.9 ± 2.0	68.0 ± 7.6	83.4 ± 3.0	72.6 ± 4.7	80.4 ± 2.7
	cv5	75.1 ± 4.3	83.9±3.3	66.9 ± 5.2	83.3 ± 5.4	71.6 ± 4.0	80.2 ± 5.5
	Cv10	74.4 ± 2.1	84.0 ± 2.7	66.3 ± 5.3	82.6 ± 3.6	71.1 ± 2.8	79.3 ± 2.9
CURVE2 MAN ALL	cv-all	77.0 ± 3.6	85.9 ± 2.9	69.3 ± 5.8	84.5 ± 4.0	73.5 ± 3.9	82.0 ± 4.2
	cv1	76.6 ± 4.8	85.6 ± 3.4	69.3 ± 6.5	84.0 ± 5.4	73.4 ± 4.8	81.4 ± 5.7
	cv5	76.6 ± 3.7	85.6 ± 2.7	68.9 ± 6.2	84.3 ± 3.0	73.2 ± 4.0	81.4 ± 3.4
	Cv10	76.4 ± 3.2	85.6 ± 2.6	68.1 ± 3.8	84.7 ± 3.6	72.7 ± 2.9	81.7 ± 4.1
RANK2 MAN ALL	cv-all	72.9 ± 3.8	83.5 ± 3.1	59.4 ± 6.0	86.7 ± 4.1	68.1 ± 3.2	81.7 ± 4.7
	cv1	73.4 ± 4.0	83.6 ± 3.0	60.0 ± 6.4	87.0 ± 4.7	68.5 ± 3.8	82.3 ± 5.3
	cv5	73.2 ± 3.4	84.1 ± 2.5	59.0 ± 6.5	87.4 ± 3.0	68.2 ± 3.4	82.5 ± 3.6
	Cv10	73.0 ± 3.3	83.4 ± 2.9	59.4 ± 6.0	86.6 ± 4.1	68.2 ± 3.2	81.7 ± 4.7

The classification performances with simple size and diameter features are also summarized in the rows denoted as MORPH. The notations “MAN”, “DRLSE”, and “GC” suggest the usage of experts’ drawings and image segmentation results from the DRLSE and GC methods, respectively, in the experiments. The rows “SDAE1”, “CURVE1”, and “RANK1” report the performance statistics of using SINGLE strategy for each algorithm, whereas the performances statistics in the rows of “SDAE2”, “CURVE2”, and “RANK2” are the results with ALL strategy in the training of each algorithm, “AUC”, “ACC”, “SENS”, “SPEC”, “PPV”, and “NPV” represents six assessment metrics: area under receiver operating characteristic curve, accuracy, sensitivity, specificity, positive predictive value, and negative predictive value, respectively. The rows “cv-all” represents the performance of each algorithm over all 100 folds, whereas the rows “cv1”, “cv5”, and “cv10” list the first, fifth, and tenth cross validations sorted by the “ACC” values of the SDAE algorithm.

Back to article page

Table 1 Performance summary of the SDAE, CURVE and RANK algorithms on the lung CT dataset.

Search

Quick links