Method: automatic segmentation of mitochondria utilizing patch classification, contour pair classification, and automatically seeded level sets

Table 1 Test results

Test #	Tissue	TPR	FPR (RLF)	FPR (CP)	Accuracy (RLF)	Accuracy (CP)
1	Cerebellum	0.8	0.05	0.02	0.94	0.97
2	Cerebellum	0.9	0.11	0.03	0.89	0.96
3	Dentate Gyrus	0.84	0.29	0.02	0.72	0.97
4	CA3	0.9	0.3	0.07	0.72	0.93

Results for all tests. TPR stands for true positive rate. FPR stands for false positive rate. Accuracy was calculated by comparing against manually labeled data. TPR applies to both the RLF and CP methods. For comparison, in each test, the RLF threshold was set so that its TPR would be equal to the TPR of the Cytoseg process. Note that CP consistently gives higher accuracy and lower FPR.

ISSN: 1471-2105