Implementation of the CNN-SVM Hybrid Model for Leukocoria Detection

Retinoblastoma (Rb) is a tumor that is a primary intraocular malignancy of the retina that generally affects infants and children. The most common symptom of Rb is a white reflection emitted from the retina of the eye through the pupil called leukocoria. This disease must be treated earlier, otherwise, it can lead to dangerous diseases. Based on this, an early diagnosis will be very important, and this will be done by detecting leukocoria. With the development of technology, image classification has become easier. Convolutional Neural Network (CNN) is an advanced image processing algorithm, and Support Vector Machine (SVM) is an advanced algorithm for general classification. For this reason, the authors propose a CNN-SVM hybrid method to classify eyes with leukocoria. The model with the CNN-SVM hybrid method will extract features from the eye photographic image on the CNN layer and perform the final classification using the SVM algorithm. The model has good performance with accuracy, precision, recall, and F1-score values reaching 96.2%, 90.9%, 100.0%, and 95.2%.