Dyspepsia is a condition of indigestion and became one of the diseases with a large number of patients in Indonesia. Early detection of Dyspepsia is done to assist in the prevention of the disease. Iridology is a method of early detection in human organs disorders by analyzing the iris patterns. However, the application of Iridology technique often finds difficulties because it requires a high level of precision in the iris image observation. The low quality of the image also leads to the high possibility of human error. Therefore, research in this paper is aimed to build Iridology-based image processing system to detect early Dyspepsia. Stages of feature extraction and classification is important and determines how the performance of the established system. The method of Linear Discriminant Analysis (LDA) is used in the feature extraction stage to reduce the image feature dimension and obtain the features vector of the image that is being observed. Meanwhile, Cascade Correlation Neural Network (CC-NN) is a classification model that is used to determine whether an observed image shows the symptoms of Dyspepsia or not. Based on previous studies, it was shown that the CC-NN with its learning mechanism capable of producing high accuracy in classification problems. With the combination of these two methods, the system can generate fairly high accuracy in detecting Dyspepsia using Iridology-based techniques. The highest accuracy rate that can be achieved by the system is 95.45% for both in training and testing set.