ICNN-Ensemble: An Improved Convolutional Neural Network Ensemble Model for Medical Image Classification

Deep learning (DL) classification has become a major research topic in the areas of cancer prediction, image cell classification, and medicine. Furthermore, DL is core other subfields. Owing to various forms ensemble models, models have achieved state-of-the-art performances fields such as However, existing cannot solve problem generalization perfectly proposed solutions only for tasks with specific datasets. Most presented their results ImageNet dataset, elaborate on insights dataset. Nonetheless, model architectures or pretrained provide same accurate datasets different classes than ImageNet. Hence, this proposes an improved convolutional neural network (ICNN-Ensemble) based representation high-resolution channels (RHRIC) systematic dropout (SMDE). ICNN-Ensemble exploits after applying RHRIC RGB images original forms, which accesses more residual feature connections represents insight into channels. SMDE applied choose members, considering changes prediction field (APF) model. In addition, executes ensembling during test set allows be trained larger batches compared model’s final training, allowing maximal effective usage graphics processing unit (GPU). Despite small size model, benchmarking Malaria dataset clearly illustrated that significantly base models.