Positive and negative HEp-2 image classification fusing global and local features

Zhou, Jiancan; Li, Yuexiang; Zhou, Xiande; Shen, Linlin

doi:10.1109/cisp-bmei.2017.8302196

Cited by 1 publication

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While, from the comparison in accuracy it is verified that our method turns out to be the second best. However, it should be noted that while Zhou et al [34] aimed to maximize accuracy, the AUC was maximized in this work. For a more direct comparison with other methods that based their optimization on accuracy, we repeated the analysis using the latter index as a figure of merit, obtaining for the best configuration (also, this time with the SqueezeNet network), the result of 94.32% accuracy and an AUC equal to 98.34%.…”

Section: Resultsmentioning

confidence: 90%

“…The best configuration analyzed was compared with other state-of-the-art methods demonstrating an excellent ability to classify the fluorescence intensity in HEp-2 images. Unfortunately, in [34] a private database was used, so a more direct comparison on the same data is not possible. The difference in accuracy between our method and that of Zhou et al, even after optimization in accuracy, is still significant.…”

Section: Discussionmentioning

confidence: 99%

“…The difference in accuracy between our method and that of Zhou et al, even after optimization in accuracy, is still significant. Perhaps the best performance obtained in [34] is not due to a lower capacity of the CNN fine-tuning, but rather (in addition to the different database used) to the fact that, in [34], a segmentation of the images was performed in search of the cells, and starting from these, a clustering was performed that led to image classification. In this way, the method, effectively eliminating the background, shows greater robustness and, therefore, better performance.…”

Section: Discussionmentioning

confidence: 99%

“…In the work of Zhou et al [34] the authors presented a fluorescence intensity classification method in which private databases are analyzed. The method makes use of the fusion of global and local type features.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Classification of (six) staining patterns. Zhou [34] -An adaptive local thresholding was applied to cell image segmentation. Two step classification: (1) with global features based on mean and variance of each channel RGB; (2) with local features based on SIFT and bag of words for doubt cases.…”

Section: Classification By K-means Clusteringmentioning

confidence: 99%

See 4 more Smart Citations

Performance of Fine-Tuning Convolutional Neural Networks for HEp-2 Image Classification

et al. 2020

View full text Add to dashboard Cite

The search for anti-nucleus antibodies (ANA) represents a fundamental step in the diagnosis of autoimmune diseases. The test considered the gold standard for ANA research is indirect immunofluorescence (IIF). The best substrate for ANA detection is provided by Human Epithelial type 2 (HEp-2) cells. The first phase of HEp-2 type image analysis involves the classification of fluorescence intensity in the positive/negative classes. However, the analysis of IIF images is difficult to perform and particularly dependent on the experience of the immunologist. For this reason, the interest of the scientific community in finding relevant technological solutions to the problem has been high. Deep learning, and in particular the Convolutional Neural Networks (CNNs), have demonstrated their effectiveness in the classification of biomedical images. In this work the efficacy of the CNN fine-tuning method applied to the problem of classification of fluorescence intensity in HEp-2 images was investigated. For this purpose, four of the best known pre-trained networks were analyzed (AlexNet, SqueezeNet, ResNet18, GoogLeNet). The classifying power of CNN was investigated with different training modalities; three levels of freezing weights and scratch. Performance analysis was conducted, in terms of area under the ROC (Receiver Operating Characteristic) curve (AUC) and accuracy, using a public database. The best result achieved an AUC equal to 98.6% and an accuracy of 93.9%, demonstrating an excellent ability to discriminate between the positive/negative fluorescence classes. For an effective performance comparison, the fine-tuning mode was compared to those in which CNNs are used as feature extractors, and the best configuration found was compared with other state-of-the-art works.

show abstract

Section: Resultsmentioning

confidence: 90%

Section: Discussionmentioning

confidence: 99%