Multilevel Strip Pooling-Based Convolutional Neural Network for the Classification of Carotid Plaque Echogenicity
Table 4
Precision and F1-score comparisons among VGG16, SPP-based VGGNet, and MSP-based VGGNet.
Fold
Methods
Performance evaluation (%)
1
VGG16
89.1
82.5
83.3
85.0
85.7
77.7
87.7
83.9
SPP-VGG
93.0
84.5
93.7
90.4
94.3
87.7
90.4
90.8
MSP-VGG
94.4
88.2
93.8
92.1
96.5
89.6
91.7
92.6
2
VGG16
93.2
73.2
86.0
84.2
88.7
70.3
89.2
82.9
SPP-VGG
92.8
89.3
94.7
92.3
95.5
88.2
94.0
92.6
MSP-VGG
94.1
88.6
95.9
92.9
96.2
89.7
94.3
93.5
3
VGG16
96.2
69.2
81.6
82.4
81.6
67.9
87.50
79.7
SPP-VGG
94.3
80.5
86.4
87.1
93.0
78.5
88.2
86.5
MSP-VGG
98.6
86.3
90.4
91.7
96.5
85.7
91.6
91.3
4
VGG16
96.3
76.9
82.9
85.4
86.7
70.4
89.0
82.6
SPP-VGG
92.5
91.3
91.8
91.9
93.2
86.3
93.9
91.1
MSP-VGG
95.5
88.7
92.4
92.2
95.5
85.1
94.2
91.6
5
VGG16
98.0
61.9
80.6
80.2
83.8
62.7
84.6
77.6
SPP-VGG
92.7
82.4
91.5
88.8
93.3
83.8
90.2
89.1
MSP-VGG
95.4
84.9
93.7
91.3
93.9
86.9
93.1
91.3
VGG16
SPP-VGG
MSP-VGG
SPP-VGG and MSP-VGG are short for SPP-based VGGNet and MSP-based VGGNet, respectively. The best results are highlighted in bold. The listed metrics were obtained on the test dataset.
