Background/Aim: Tomography imaging is a valuable alternative to colonoscopy for diagnosing colorectal cancer cases especially when colonoscopy is not applicable. Therefore, processing tomography images by computer-aided diagnosis support systems is essential for helping clinicians. The aim of this study is to investigate the applicability of convolutional neural networks (CNN) for diagnosis of colorectal cancer from tomography images. This study uses CNN models to classify abdominal tomography im Methods: ages automatically. The image classication is performed using four benchmark CNN models, namely, MobileNETv2, ResNet101V2, InceptionResnetV2, EfcientNetB2, and a shallower sequential CNN model. The benchmark models are experimented with using transfer learning for two separate cases: with and without data augmentation. According to the results, the proposed sequential CNN model achieve Results: d the highest classication scores with mean accuracy, precision, recall and F1-score of 0.9803, 0.9813, 0.9793, and 0.9803, respectively. It is Conclusion: possible to achieve a better detection performance when a shallower CNN model with lower number of parameters is used. Making prediction with such a model means performing less computation to achieve the result. All in all, computer aided diagnosis of colorectal cancer is possible to help the clinicians during their daily practices