Human intestinal parasites are the main cause of major diseases and health problems in children and adults, but the most common occurrence is in children. Helminth is one of the intestinal parasitic worms that has caused parasitic worm disease to occur, which is also known as helminthiasis disease. Symptoms of helminthiasis are anaemia, diarrhoea, and severe problems of malnutrition. Early diagnosis of this disease is necessary for the recovery of patients, especially children. Helminth ova can be diagnosed through the patient's stool, cells, and blood sample. The observation and detection of helminth ova are conducted manually using a light microscope, and the results are prone to human error. The procedure is time-consuming and easily results in a human error if the parasitologist is not focused. Due to the requirement of a diagnosis system with high accuracy of helminth ova, this research is proposed. The main objective of this research is to provide an automated intelligence classification system for human intestinal parasites based on helminth ova that consists of image processing combined with artificial intelligence (AI) for helminth ova classification. The proposed techniques have been tested on four types of intestinal parasitic helminth ova, which are Ascaris Lumbricoides ova (ALO), Enterobius Vermicularis ova (EVO), Hookworm ova (HWO), and Trichuris Trichiura ova (TTO). Each species consists of 100 images with normal exposure, 33 images with under-exposure, and 33 images with over-exposure. The overall images used for these four helminth species are 664 images. The methodology stages of this research consist of enhancement, color model, segmentation, feature extraction, and classification. One technique has been selected for each methodology stage by comparing several techniques to obtain the best results. The sequence of the techniques chosen is Modified Global Contrast Stretching (MGCS) for enhancement stage, K component from CMYK color model for color conversion stage, fuzzy c-means for segmentation stage, and Linear kernel of support vector machine (SVM) classifier for classification stage. The best technique has been selected for enhancement based on the qualitative and quantitative analysis, which the MGCS technique has obtained compared to the other enhancement techniques. Then, for the color model, the S component from the HSV color space has obtained the highest accuracy, specificity, and intersection over union (IoU) analysis with values of 98.99%, 72.83%, and 53.42%, respectively. Meanwhile, the K component from the CMYK color model has obtained the highest sensitivity with a value of 79.78%. In the segmentation stage, fuzzy c-means obtained the highest results with 98.43% for accuracy, 85.05% for sensitivity, 98.64% for specificity, and 56.37% for IoU using the K component from the CMYK color model. The classification accuracy obtained for the SVM classifier with Linear kernel is the highest value, 92.23%, compared to Ensemble classifier through AdaBoostM2 aggregation with accuracy, 89.94%, and k-Nearest Neighbourhood (k-NN) classifier through Cityblock distance of accuracy, 91.16%.
