Keynote Speakers
Opportunities and challenges in AI for Endoscopy/Surgery
PhD. Sharib Ali
I am an assistant professor at the School of Computing, University of Leeds in the United Kingdom. Before, I was a post-doctoral researcher at the German Cancer Research Centre in Germany and at the University of Oxford in the United Kingdom. My research expertise is in biomedical image analysis focusing on endoscopic and surgical computer vision. Currently I am leading research group in Artificial Intelligence for Medicine and Surgery (AIMS group, https://artificial-intelligence.leeds.ac.uk/aims/).
I have published over 20 peer-reviewed journals (including medical image analysis, IEEE TNNLS, IEEE JBHI, Neuroimage, Pattern Recognition, CVIU, Gastroenterology, and Cancer Research), over 45 peer-reviewed conference papers (MICCAI, IEEE ISBI, CVPR and others), and two patents. I am also the initiator, lead organiser and chair of the endoscopic computer vision challenge series started in 2019 in conjunction with IEEE International Symposium on Biomedical Imaging (EndoCV @ IEEE ISBI). Recently, I initiated another challenge on pre-operative to intraoperative fusion at a Medical Image Computing and Computer-Assisted Intervention conference (P2ILF @ MICCAI2022). These ventures are joint effort between clinical and computational scientists from various universities and hospitals. I am the organiser and chair of Cancer Prevention through Early Detection workshop series launched in 2022, and a newly launched date engineering in medical imaging workshop in 2023 at Medical Image Computing and Computer-Assisted Intervention conference (MICCAI).
Opportunities and challenges in AI for Endoscopy/Surgery
Whilst current data-driven approaches have demonstrated the ability to democratise digital healthcare in general, it has tremendous opportunities in procedures that depend on endoscopic and surgical systems. One main reason is the high operator dependence that requires envisioning organs and performing procedures through continuous navigation, monitoring and interactions. This is the utmost challenge and requires special skills, coordination, and expertise. In addition, the challenges in understanding the photogenic images of a lesion to be cancerous or not are also subjective and require more comprehensive learning.
During surgical procedures in minimally invasive surgery, knowledge about the surrounding tissues, organs and blood vessels is essential. Still, often, this can be a surplus of information to process by a single surgeon’s brain and experience. Often, these procedures are also limited in field-of-view and suffer from visual occlusions, organ movements and quality of images in general due to the interaction of light with fluids. These challenges are common in almost all endoscopic procedures independent of organ type or disease, from recognising the subtle dysplasia in a healthy mucosal pattern to finding precancerous lesions in the surrounding inflammatory tissue. While resections must be perfect for complete recovery, there is no objective means to quantify them. Longitudinal monitoring of patients is still a myth. Artificial intelligence (AI)-driven methods are being produced in academic and industrial settings. However, there is a massive gap in filling all these challenges and building comprehensive and generalisable methods that can be adapted to various populations and clinical settings.
In this talk, I will walk through my journey of involving deep learning approaches to answer some of these questions and my efforts in tackling some of these challenges. I will also highlight my current focus on involving multimodality data and efforts to minimise biases. Visualisation is another critical component in this field - from mapping the oesophagus to the colon to applying augmented reality in minimally invasive surgery.
Organizing Institutions
