Recognition and Detection of Mitosis Event Based on Feature of Evolution in Time Domain

Chuang Chen; Wenwu Jia; Ya Wang

doi:10.3788/LOP56.241007

Journals >Laser & Optoelectronics Progress >Volume 56 >Issue 24 >Page 241007 > Article

Laser & Optoelectronics Progress
Vol. 56, Issue 24, 241007 (2019)

Recognition and Detection of Mitosis Event Based on Feature of Evolution in Time Domain

Chuang Chen, Wenwu Jia^*, and Ya Wang^**

Author Affiliations

School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China

show less

DOI: 10.3788/LOP56.241007 Cite this Article Set citation alerts

Chuang Chen, Wenwu Jia, Ya Wang. Recognition and Detection of Mitosis Event Based on Feature of Evolution in Time Domain[J]. Laser & Optoelectronics Progress, 2019, 56(24): 241007 Copy Citation Text

show less

Fig. 1. Processes of mitosis recognition and detection

Download full size

Fig. 2. Whole process of mitosis. It means that mitosis is completed when shape of “8” appears

Download full size

Fig. 3. Results of mitosis locations in phase-contrast microscopy image on C2C12 dataset. (a) Detection result of 918th frame; (b) detection result of 952nd frame; (c) detection result of 782nd frame; (d) detection result of 902nd frame

Download full size

Method	Accuracy /%	Precision /%	Recall /%	F-score /%
Level-1	97.89	95.78	98.20	96.97
Level-2	97.95	98.14	95.91	96.97
Combine	98.30	97.81	97.27	97.51

Table 1. Experimental results with different levels on C2C12 dataset

Pooling	Accuracy /%	Precision /%	Recall /%	F-score /%
Sum pooling	96.53	94.22	95.93	95.05
Max pooling	98.46	97.66	97.89	97.76
Gradient pooling 1	98.44	97.83	97.66	97.73
Gradient pooling 2	98.14	95.69	99.13	97.39
Max+Sum+G1	98.69	98.05	98.22	98.12
G1+G2	98.61	97.64	98.37	97.97
Max+Sum+G1+G2	98.30	97.82	97.27	97.51

Table 2. Experimental results with different pooling strategies on C2C12 dataset

Feature	Accuracy /%	Precision /%	Recall /%	F-score /%
GIST	98.30	97.82	97.27	97.51
SIFT	98.27	97.34	97.25	97.29
CNN	98.34	96.34	99.04	97.67

Table 3. Experimental results with different features on C2C12 dataset

Feature	Accuracy /%	Precision /%	Recall /%	F-score /%
GIST	89.92	93.32	88.91	91.61
SIFT	88.96	93.53	88.96	91.18
CNN	95.77	94.70	89.40	94.95

Table 4. Experimental results with different features on C3H10 dataset

Method	F-score /%	Precision /%	Recall /%
EOF+SVM	97.51	97.82	97.27
BoW+SVM	87.80	95.20	91.30
HCRF	91.80	91.97	92.50
HSCRF	93.12	92.40	93.70
HSCNF	93.50	92.40	94.60

Table 5. Comparison of EOF+SVM with other methods on C2C12 dataset

Method	F-score /%	Precision /%	Recall /%
EOF+SVM	94.95	94.70	89.40
MM-HCRF+MM-SMM^[18]	91.80	95.80	88.10
MM-HCRF^[18]	87.20	82.80	92.20
EDCRF^[20]	88.90	91.30	87.00
CRF^[8]	81.50	90.50	75.30
HMM^[6]	81.00	83.40	79.40

Table 6. Comparison of EOF+SVM with other methods on C3H10 dataset

Chuang Chen, Wenwu Jia, Ya Wang. Recognition and Detection of Mitosis Event Based on Feature of Evolution in Time Domain[J]. Laser & Optoelectronics Progress, 2019, 56(24): 241007

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information