This repository implements a machine learning pipeline using Active Learning modAL package and Keras to classify endoscopic video capsule images into informative or non-informative. By using Active Learning, we aim to reduce the costs of annotation.
Machine learning plays a crucial role in detecting informative images in video capsule endoscopy. Video capsule endoscopy is a non-invasive diagnostic method used to examine the inside of the gastrointestinal tract. During this procedure, a small capsule equipped with a camera is swallowed, and the resulting images are used to identify and diagnose various gastrointestinal conditions. However, the vast amount of data generated by video capsule endoscopy can be overwhelming for human annotators to review, making it a time-consuming and labor-intensive task.
By using machine learning algorithms to classify images as informative or non-informative, healthcare professionals can quickly and accurately identify relevant images, reducing the time and effort required to review the data. Additionally, machine learning models can be trained to identify specific patterns and characteristics in the data that may be indicative of certain conditions, increasing the accuracy of diagnosis. With the ability to detect informative images more efficiently and accurately, machine learning has the potential to greatly improve the effectiveness and efficiency of video capsule endoscopy as a diagnostic tool.
This implementation uses an Active Learning strategy, where an initial small labeled dataset is used to train the model, and then the model is used to predict the labels of the remaining unlabeled data. The samples with the highest uncertainty are then selected and annotated by a human annotator, and the labeled data is added back to the training set to fine-tune the model. This process is repeated until the desired performance is achieved or the budget for annotation is exhausted.
By using Active Learning, we can reduce the amount of annotated data needed to train a model to a high level of accuracy, making the process more cost-effective and efficient.