An Ensemble of Machine Learning Models Utilizing Deep Convolutional Features for Medical Image Classification

Medical image classification is a rapidly growing research field that has revolutionised various diseases' traditional diagnosis, treatment planning, and prognosis prediction. Due to the recent advancements in deep learning (DL) algorithms for medical imaging, the ability to identify anomalies and crucial features in medical images has greatly improved, enhancing the precision of diagnosis and treatment. However, the training process of deep learning algorithms is time-consuming and resource-intensive. In contrast, a lightweight model can be easily deployed for real-time implementation. Furthermore, in preceding studies, single models are primarily deployed for performing medical image classification tasks, though an ensemble of various models can enhance overall performance. Therefore, in this work, an ensemble of six machine learning (ML) models is used to reduce the training time compared to a conventional DL model. Moreover, the features considered in this work are deep convolutional features extracted using pre-trained DL models. The dataset considered in this work includes X-rays, CT scans, and MRI images of human body parts associated with several diseases. Our proposed ensemble learning approach is experimentally proven to outperform individual models, considering the deep convolutional features.