MRI BRAIN TUMOR CAPTIONING USING VGG IMAGE ANNOTATOR

Abstract Digital medical images have become increasingly important in the last several years for the detection of more disorders. The health care industry needs an accurate and efficient AI diagnostic system for the compilation of medical reports. This aims for object detection technique, which uses the computer vision (CV) and deep learning(DL) based pre trained Convolutional Neural Networks (CNNs) architectures to extract more accurate features and to classify the objects in the MR Brain images. Natural language processing, and Recurrent Neural Networks (RNNs) techniques were utilized to complete the task of generating high quality captions from these identified features. Automatic image captioning has made a significant progress in recent years, and it is more difficult problem. In order to implement NLP into a medical image captioning system using DL techniques, certain image annotation tools are available as an open source tools. Some popular AI image caption generator tools available which includes LabelImg, MakeSense.ai, SuperAnnotate, CVAT (Computer Vision Annotation Tool) etc. Visual Geometry Group Image Annotator tool (VGGIA), is one of the effective tools available in open source, which helps to define the caption for the regions affected in the MR Brain images. This paper explores the usage of VGGIA for captioning brain tumor image to illustrate NLP approach for deep learning based models. Keywords : Deep learning (DL), Natural Language Processing (NLP), Magnetic Resonance Images (MRI), Visual Geometry Group Image Annotator tool (VGGIA), Artificial Intelligence (AI).