Comparison of CNNs and SVM for Detection of Activation in Malaria Cell Images
Autor: | Jale Bektaş |
---|---|
Rok vydání: | 2019 |
Předmět: |
Computer Science
Information System classification CNN SVM Pattern recognition business.industry Computer science Cell Pattern recognition General Medicine medicine.disease Support vector machine medicine.anatomical_structure Pattern recognition (psychology) medicine Bilgisayar Bilimleri Bilgi Sistemleri Artificial intelligence business Malaria |
Zdroj: | Volume: 2, Issue: 2 38-50 Natural and Applied Sciences Journal |
ISSN: | 2645-9000 |
Popis: | Malaria is a disease caused by parasites that are transmitted through the enzymes of Anophele mosquito and cause symptoms in fatal danger. Thick and thin film microscopic examination of smears taken from blood is the most reliable method for diagnosis. In the manual examination of the smears, the expertise of examiner and the quality of the smear significantly affect the accuracy of the diagnosis. Malaria's automatic diagnosis of pattern recognition and classification techniques on blood smear images is among the subjects of research. In this study, well-known Convolutional Neural Networks include InceptionV3, GoogLeNet, AlexNet, Resnet50, Vgg16 networks and six-fold cross validation was applied and performance evaluations were performed with a Machine Lerning method, Support Vector Machine. It was found that Deep Learning methods achieved at least 10.08% of accuracy difference performance compared to SVM based on the features of the input sample images. This difference has been 0.07 for F-Score, 0.06 for sensitivity. |
Databáze: | OpenAIRE |
Externí odkaz: |