Artificial Intelligence Based System and Method for Data Augmentation in Biomedical Computer Vision Tasks

Deep learning (DL) has gained significant traction in computer vision tasks, particularly within the healthcare domain. Medical image analysis using DL models for disease segmentation and classification often requires vast amounts of annotated data. Nevertheless, obtaining annotated data can be a laborious and costly process. Data augmentation techniques address limited annotated data, overfitting, and class imbalance issues during DL model training. While existing libraries facilitate code-based data augmentation, we present a novel graphical user interface (GUI) designed to streamline this process for researchers. This user-friendly GUI allows intuitive selection of data augmentation techniques tailored to specific datasets and applications. The GUI integrates the Albumentation library and leverages Google Colab for cloud-based processing. Our implementation demonstrates successful data augmentation on medical images, generating 92 augmented versions from a single input image, significantly reducing researcher effort in data preparation.