Deep convolutional neural networks (CNN) manifest the potential for computer-aided diagnosis systems (CADs) by learning features directly from images rather than using traditional feature extraction methods. Nevertheless, due to the limited sample sizes and heterogeneity in tumor presentation in medical images, CNN models suffer from training issues, including training from scratch, which leads to overfitting. Alternatively, a pre-trained neural network’s transfer learning (TL) is used to derive tumor knowledge from medical image datasets using CNN that were designed for non-medical activations, alleviating the need for large datasets. This study proposes two ensemble learning techniques: E-CNN (product rule) and E-CNN (majority voting). These techniques are based on the adaptation of the pretrained CNN models to classify colon cancer histopathology images into various classes. In these ensembles, the individuals are, initially, constructed by adapting pretrained DenseNet121, MobileNetV2, InceptionV3, and VGG16 models. The adaptation of these models is based on a block-wise fine-tuning policy, in which a set of dense and dropout layers of these pretrained models is joined to explore the variation in the histology images. Then, the models’ decisions are fused via product rule and majority voting aggregation methods. The proposed model was validated against the standard pretrained models and the most recent works on two publicly available benchmark colon histopathological image datasets: Stoean (357 images) and Kather colorectal histology (5,000 images). The results were 97.20% and 91.28% accurate, respectively. The achieved results outperformed the state-of-the-art studies and confirmed that the proposed E-CNNs could be extended to be used in various medical image applications.

The rapid expansion of medical data poses numerous challenges for Machine Learning (ML) tasks due to their potential to include excessive noisy, irrelevant, and redundant features. As a result, it is critical to pick the most pertinent features for the classification task, which is referred to as Feature Selection (FS). Among the FS approaches, wrapper methods are designed to select the most appropriate subset of features. In this study, two intelligent wrapper FS approaches are implemented using a new meta-heuristic algorithm called Sand Cat Swarm Optimizer (SCSO). First, the binary version of SCSO, known as BSCSO, is constructed by utilizing the S-shaped transform function to effectively manage the binary nature in the FS domain. However, the BSCSO suffers from a poor search strategy because it has no internal memory to maintain the best location. Thus, it will converge very quickly to the local optimum. Therefore, the second proposed FS method is devoted to formulating an enhanced BSCSO called Binary Memory-based SCSO (BMSCSO). It has integrated a memory-based strategy into the position updating process of the SCSO to exploit and further preserve the best solutions. Twenty one benchmark disease datasets were used to implement and evaluate the two improved FS methods, BSCSO and BMSCSO. As per the results, BMSCSO acted better than BSCSO in terms of fitness values, accuracy, and number of selected features. Based on the obtained results, BMSCSO as a FS method can efficiently explore the feature domain for the optimal feature set.