Human-in-the-Loop Deep Learning for Nucleus Detection in Oral Cancer Screening – Olle Edgren Schüllerqvist

Despite early detection being widely recognized as a key strategy to reduce the global burden of oral cancer, there are currently no widely adopted oral cancer screening programs. Advancements in artificial intelligence and deep learning over the last decade have enabled new possibilities in medical imaging, and a deep learning-based pipeline for oral cancer screening based on cytological whole-slide images has been proposed.

This seminar is based on my recently published master’s thesis, which investigates whether an active learning-based human-in-the-loop approach can improve the performance of deep learning models for nucleus detection in a label-efficient manner. Considering both theoretical and practical aspects of label efficiency, we will examine three active learning methods and a practical workflow for interactive annotation with large-scale applications such as mass screening in mind.

In particular, this seminar will focus on the computational methods developed, the main findings of two experiments, and my experience designing and implementing solutions for efficient labeling integrated with the active learning process. Experimental results demonstrate that active learning can significantly improve the label efficiency of nucleus detection compared to a baseline method of random sampling. In combination with the developed infrastructure, this work highlights the potential impact of active learning-based human-in-the-loop solutions for large-scale clinical applications.