IoT-enabled image captioning with deep learning for healthcare domain

Abstract

<p>The ever-increasing volume of medical images greatly strains clinicians who are in the process of reviewing it and writing reports. It would be more efficient and cost-effective if an image captioning model could automatically create report drafts from matching photos, thereby relieving physicians from this tedious work. The Internet of things (IoT) has switched its emphasis from its initial binary concept to that of the Internet of multimedia things (IoMT) because of the explosive rise of multilingual-on-demand data in various sound, footage, picture forms. This work proposed a deep learning-based image caption network (DL-ICN) for healthcare domain. The work originality is shown using DL to identify various class labels of the patient X-ray and ECG images. With the help of bilateral encoder representations from transformers (BERT) method for captioning pictures, a detailed written summary of a person’s medical picture may be generated automatically. Results of simulations showed that the proposed model achieved good compression performance, good quality reconstruction and good classification results for image captioning.</p>