Advancing Education, Research, and Quality of Care for the Head and Neck oncology patient.
Introduction: Head and neck cancer poses a significant global health burden. The limited availability of skilled subspecialty surgeons highlights the potential role of computer vision (CV) and artificial intelligence (AI) in assisting both novice and experienced surgeons. These technologies can help with procedural navigation and enhance diagnostic accuracy in clinical practice.
Methods: A systematic review following PRISMA guidelines was conducted across PubMed, Elsevier, and Medline databases, covering studies on computer vision in head and neck Pathology (HNP) from 2019 to 2024.
Results: A total of 27 studies were identified: 25 focused on photographic analysis, while 2 focused on video analysis. Retrospective image analysis was the focus of 24 studies, with 3 addressing real-time image processing. Traditional machine learning (ML) methods were employed in 2 studies (7%), while 25 studies (93%) utilized deep learning (DL) techniques. 21 models were supervised learning, 3 were unsupervised and 2 were semi-supervised. Reported sensitivity, specificity, and precision for DL models ranged from 91.7%-75.0%, 98.75%-75.0%, and 99.0%-71.0%, respectively.
Discussion: Most studies demonstrated a trade-off between accuracy and speed, with a marked decrease in performance when attempting real-time analysis. Novel unsupervised deep learning approaches aim to address this challenge. However, the review emphasizes the need for less resource-intensive models and larger publicly available HNP datasets for training to improve model performance and scalability.
Conclusion: The use of computer vision in identifying HNPs and tracking surgical procedures is increasing, particularly in retrospective analysis. Supervised deep learning and convolutional neural networks are being used more extensively to improve specificity, sensitivity, and accuracy in a time and power-efficient manner.
Keywords: Artificial Intelligence (AI), Computer vision, Deep Learning, Diagnostic, Head and Neck Cancer, Machine learning, Neural Networks, Oral Cancer