Medical industry is one of the well-known sectors where predictions are very essential. This research study encourages the use artificial intelligence for decision-making process in medical field for detecting acute appendicitis disease. Acute appendicitis is one of the most common abdominal pain that leads to emergency surgery. Currently, ultrasound has gained high acceptance for diagnosis of patients with severe abdominal pain because it can have no any radiation, cost effective, and is easily available. The main objective of this study is to develop a Convolutional Neural Network Based Detection System for Acute Appendicitis Disease. The existing acute appendicitis diagnosis systems mainly focus on the clinical and laboratory findings. But now a day, these results lacks acceptance due to they result in false diagnosis. To minimize this, results from imaging modalities have gained high acceptance worldwide. To help the physicians in decision making process for diagnosis acute appendicitis from imaging modalities specifically the ultrasound images, an intelligent model that can diagnosis acute appendicitis from ultrasound image is very important. The image datasets which have been used for this study were collected from internationally available sources (pediatrics, 2011) (appendicitis during pregnancy, 2013) (encyclopaedia, n.d.) (abdomen and retroperitoneum, 2012) (Infant with perforated appendicitis , 2014). The Gaussian filter and histogram equalization techniques were applied in order to remove the noises in the image and increase the brightness of the images respectively. The proposed model is trained and validated for both pre-processed and unprocessed data. The model is tested with test data and the results obtained are used to evaluate the performance of the model. The performance of the proposed model is evaluated based accuracy, recall and specificity and the results obtained from the experimental analysis were presented. The proposed model achieved an accuracy of 99.16 for pre-processed data and 95.18 for unprocessed data. The performance of the proposed model is compared with pre-trained models and the result shows that our proposed model is a good choice than pre-trained ones.