Toward High-Throughput Artificial Intelligence-Based Segmentation in Oncological PET Imaging

TL;DR

This paper reviews AI-based segmentation methods in oncological PET imaging, emphasizing the need for high-throughput, semi-supervised, and unsupervised techniques to facilitate clinical adoption amid limited annotated data.

Contribution

It provides a comprehensive review of existing AI segmentation techniques and evaluation criteria for clinical translation in PET imaging.

Findings

AI techniques have advanced segmentation accuracy in PET imaging.

Semi-supervised and unsupervised methods address data annotation challenges.

Evaluation criteria are crucial for translating AI methods into clinical practice.

Abstract

Artificial intelligence (AI) techniques for image-based segmentation have garnered much attention in recent years. Convolutional neural networks (CNNs) have shown impressive results and potential towards fully automated segmentation in medical imaging, and particularly PET imaging. To cope with the limited access to annotated data needed in supervised AI methods, given tedious and prone-to-error manual delineations, semi-supervised and unsupervised AI techniques have also been explored for segmentation of tumors or normal organs in single and bi-modality scans. This work provides a review of existing AI techniques for segmentation tasks and the evaluation criteria for translational AI-based segmentation efforts towards routine adoption in clinical workflows.