Automated Medical Report Generation using Deep Learning and Attention Mechanism

One of the essential yet challenging tasks in medical diagnostics is the automatic production of well-informed reports from medical images. Radiology has traditionally been the specialty that reads all types of medical images, but report writing by hand is prone to mistakes, time-consuming and difficult even for experts. In this paper, we investigate the ability of advanced technologies such as Deep Learning and attention schemes in scripting medical imaging reports from scratch. With this extensive research, under the backdrop of the recent works, we represent methodologies and techniques for Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, co-attention mechanisms as well hierarchical LSTM models to tackle the challenges in medical imaging interpretation. Here we embark on the details and subtleties of this space: from detecting anomalies to writing a systematic report. An analysis of literature and experimentation forms the backbone of this paper, and the aim is to show the current trends in automatic medical image report generation, as an ongoing challenge that transpired from the rapid growth witnessed in this transformative area.