Desktop version

Home arrow Engineering arrow Cellular Image Classification

Experimental Results on the ICPR2012 Dataset

We first evaluate our proposed AdaCoDT method on the ICPR2012 dataset. We choose linear SVM classifier due to its effectiveness and efficiency. The linear SVM is trained based on the training set by 10-fold cross validation strategy and tested using the test set. Table7.2 shows the classification performance of each method. The AdaCoDT method outperforms all the other methods, achieving 75.2 % of classification accuracy on individual cells. The obtained classification accuracy is even higher than that of a human expert. It is worth noting that the AdaCoDT method significantly outperforms CoALBP [19] which is the winner of the contest.

Table7.3 illustrates the confusion matrices presenting the classification performance for each staining pattern at the cell level. It is obvious that cytoplasmic, centromere and homogeneous patterns are classified more accurately than the others. More particularly, cytoplasmic can achieve 100% of classification accuracy. However, the classification accuracy for fine speckled pattern is only 52.6 %. Compared to the cytoplasmic pattern with distinguishable shape and centromere pattern with clear fluorescent dots, speckled pattern and homogeneous pattern has similar characteristic and is hard to find discriminative features to separate.

To evaluate the classification performance at the image level, we report the corresponding confusion matrix in Table7.4. Our proposed AdaCoDT method obtains the classification accuracy of 85.7 %. Centromere, cytoplasmic, homogeneous and

Table 7.2 Classification performance on the ICPR2012 dataset

Algorithm

Average accuracy (%)

Sensitivity (%)

AdaCoDT

75.2

77.1

Human [25]

73.3

/

LSC-SIFT

68.9

70.6

LSC-CoDT

66.9

66.5

FK-SIFT

67.6

66.4

CoALBP [25]

68.7

70.4

RIC-LBP [20]

67.5

67.6

LBP [1]

58.9

59.2

Table 7.3 The confusion matrix for the cell level classification on the ICPR2012 dataset

ce (%)

cs (%)

cy (%)

fs (%)

ho (%)

nu (%)

ce

85.9

8.1

0.0

0.0

0.0

6.0

cs

4.0

75.3

2.9

17.8

0.0

0.0

cy

0.0

0.0

100.0

0.0

0.0

0.0

fs

20.2

3.5

6.2

52.6

17.5

0.0

ho

8.3

2.8

0.6

11.1

73.9

3.3

nu

2.2

0.0

11.5

3.6

7.9

74.8

Table 7.4 The confusion matrix for the image level classification on the ICPR2012 dataset

ce (%)

cs (%)

cy (%)

fs (%)

ho (%)

nu (%)

ce

100.0

0.0

0.0

0.0

0.0

0.0

cs

0.0

66.7

0.0

33.3

0.0

0.0

cy

0.0

0.0

100.0

0.0

0.0

0.0

fs

50.0

0.0

0.0

50.0

0.0

0.0

ho

0.0

0.0

0.0

0.0

100.0

0.0

nu

0.0

0.0

0.0

0.0

0.0

100.0

nucleolar patterns achieve 100 % classification accuracy. The most frequent mistake is existed between fine speckled and centromere pattern, which is common mistake at the cell level.

 
Source
< Prev   CONTENTS   Source   Next >

Related topics