Classification of chest X-ray images by incorporation of medical domain knowledge into operation branch networks

BMC Medical Imaging

Table 6 Results of attention index for true positive data

Dataset	Backbone	Conventional	ABN	OBN1	OBN2	OBN3
Teikyo	ResNet50	0.60 ± 0.11	0.59 ± 0.11	0.56 ± 0.10	0.57 ± 0.11	0.58 ± 0.10
Teikyo	DenseNet121	0.63 ± 0.11	0.65 ± 0.13	0.72 ± 0.13	0.69 ± 0.13	0.72 ± 0.11
Pulmonary hypertension	ResNet50	0.60 ± 0.25	0.57 ± 0.27	0.69 ± 0.23	0.70 ± 0.24	0.62 ± 0.28
Pulmonary hypertension	DenseNet121	0.55 ± 0.15	0.65 ± 0.16	0.68 ± 0.20	0.77 ± 0.13	0.69 ± 0.16
Heart failure	ResNet50	0.74 ± 0.25	0.73 ± 0.24	0.69 ± 0.25	0.78 ± 0.20	0.66 ± 0.23
Heart failure	DenseNet121	0.75 ± 0.21	0.56 ± 0.31	0.84 ± 0.19	0.84 ± 0.21	0.90 ± 0.09

This table presents the averaged Attention index over all ten splits, with the respective calculated standard deviations for true positive data: ABN, attention branch network; OBN1, operation branch network using weight map with a convex hull on mask images of lung field; OBN2, operation branch network using weight maps with combined mask images of the lung field and heart

ISSN: 1471-2342