Early Detection of Retinoblastoma Using K-Nearest Neighbors and HSV Color Features

Retinoblastoma (RB) is the most common malignant intraocular cancer in children and may lead to blindness or death if not detected early. Most existing detection methods utilize deep learning approaches, which require large datasets and high computational resources. Therefore, this study proposes a simple and efficient early detection method for retinoblastoma using HSV (Hue, Saturation, Value) color features and the K-Nearest Neighbors (KNN) algorithm. The dataset consisted of 105 eye images, including 52 normal eye images and 53 retinoblastoma-positive images. The research stages included region-of-interest extraction in the pupil area, conversion of images from the RGB color space to HSV, feature extraction based on average saturation and average brightness values, and classification using KNN. Model evaluation was performed using the Leave-One-Out Cross-Validation (LOOCV) method with k-value variations ranging from 1 to 10. The test results showed that k = 1 achieved the best performance, with an accuracy of 83.81%, sensitivity of 83.02%, specificity of 84.62%, precision of 84.62%, and an F1-score of 83.81%. These results demonstrated that the combination of HSV color features and KNN classification was able to effectively distinguish between normal eye images and retinoblastoma cases. The proposed method has the potential to serve as a simple and low-cost screening tool for early retinoblastoma detection.