An Improved Three-Dimensional Dual-Path Brain Tumor Image Segmentation Network

Hengliang Zhang; Qiang Li; Xin Guan

doi:10.3788/AOS202141.0310002

Journals >Acta Optica Sinica >Volume 41 >Issue 3 >Page 0310002 > Article

Acta Optica Sinica
Vol. 41, Issue 3, 0310002 (2021)

An Improved Three-Dimensional Dual-Path Brain Tumor Image Segmentation Network

Hengliang Zhang, Qiang Li^*, and Xin Guan

Author Affiliations

School of Microelectronics, Tianjin University, Tianjin 300072, China

show less

DOI: 10.3788/AOS202141.0310002 Cite this Article Set citation alerts

Hengliang Zhang, Qiang Li, Xin Guan. An Improved Three-Dimensional Dual-Path Brain Tumor Image Segmentation Network[J]. Acta Optica Sinica, 2021, 41(3): 0310002 Copy Citation Text

show less

Fig. 1. Structure of DPN-MS

Download full size

Fig. 2. Module structures of DPN-MSUint

Download full size

Fig. 3. Structure of multiplexer

Download full size

Fig. 4. Comparison diagram of three convolutions. (a) Conventional convolution; (b) group convolution; (c) group convolution with channel random mixing module

Download full size

Fig. 5. Dice box diagram on test set. (a) DMFNet; (b) DPN-MS

Download full size

Fig. 6. Graph of segmentation results. (a) Flair; (b) T1; (c) GT; (d) DPN-MS

Download full size

Fig. 7. Comparison of segmentation results. (a) Flair; (b) GT; (c) 3D-ESPNet; (d) DMFNet; (e) DPN-MS

Download full size

Penalties	D_sc /%			D_HD
Penalties	ET	WT	TC	ET	WT	TC
α=0.5,β=0.5	80.56	90.23	86.28	2.91	3.97	5.65
α=0.4,β=0.6	79.23	89.89	85.02	2.50	5.11	6.05
α=0.3,β=0.7	80.16	89.92	84.59	2.43	3.92	6.21
α=0.2,β=0.8	80.30	90.02	84.06	2.58	4.15	6.38
α=0.1,β=0.9	79.09	89.48	82.16	2.63	4.14	7.86

Table 1. Comparison of different parameters in brain tumor image segmentation

Model	P /M	O_FP /G	D_sc /%			D_HD
Model	P /M	O_FP /G	ET	WT	TC	ET	WT	TC
3D-UNet^[8]	16.21	1669.53	75.96	88.53	71.77	6.04	17.10	11.62
3D-ESPNet^[12]	3.63	76.51	73.70	88.30	81.40	-	-	-
DeepMedic^[19]	9.45	203.96	78.75	90.47	81.35	3.81	4.32	7.56
MFNet^[11]	3.19	20.61	79.91	90.43	84.61	2.68	4.68	6.31
DMFNet^[11]	3.88	27.04	80.12	90.62	84.54	3.06	4.66	6.44
NVDLMED^[20]	40.06	1495.50	81.73	90.68	86.02	3.82	4.52	6.85
DPN-MS	6.09	25.21	80.56	90.23	86.28	2.91	3.97	5.65

Table 2. Comparison of different algorithms in brain tumor image segmentation

Hengliang Zhang, Qiang Li, Xin Guan. An Improved Three-Dimensional Dual-Path Brain Tumor Image Segmentation Network[J]. Acta Optica Sinica, 2021, 41(3): 0310002

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information