Encrypted login | home

Program Information

How to Select and Evaluate a PET Auto-Segmentation Tool - Insights From AAPM TG211


C Schmidtlein

I El Naqa

A Kirov

E Spezi





C Schmidtlein1*, I El Naqa2*, A Kirov1*, E Spezi3*, (1) Mem. Sloan-Kettering Cancer Ctr, New York, NY, (2) University of Michigan, Ann Arbor, MI, (3) Memorial Sloan-Kettering Cancer Center, New York, NY, (3) Cardiff University, Cardiff, Wales UK

Presentations

7:30 AM : Understanding PET Images for Segmentation Tasks - C Schmidtlein, Presenting Author
8:00 AM : State-of-the-art of current PET-AS algorithms and their advantages and limitations for clinical application - I El Naqa, Presenting Author
8:30 AM : Components of a standard and a procedure for evaluation of PET-AS methods - A Kirov, Presenting Author
9:00 AM : Design, implementation and first results of the future standard for evaluation of PET-AS methods - E Spezi, Presenting Author

TH-AB-702-0 (Thursday, August 3, 2017) 7:30 AM - 9:30 AM Room: 702


Positron emission tomography (PET) provides information that can be used for a variety of clinical applications including patient staging and prognosis, radiation therapy planning, therapy monitoring, and the detection/prediction of recurrences or metastatic disease. Accurate delineation of the functional tumor volume in PET plays an important role for using this information in the clinical process.

AAPM Task Group No. 211 has published earlier this year an educational report summarizing the state-of-the-art in PET auto-segmentation (PET-AS) algorithms. In the process of its work the task group realized that the validation of various PET-AS algorithms is highly variable, and often inconsistent and insufficient. For this reason about one-third of the report is dedicated to the components of a standard that the task group is proposing for evaluating the PET-AS tools. A follow-up publication describes an implementation of the proposed standard into a software tool, which is intended to become publicly available. The last part of the report describes a procedure for validating PET-AS tools for clinical use, discusses the limitations and dependencies of these tools, and the need of physician verification.

This course will summarize and highlight the main findings and conclusions of the task group and some insights of how these conclusions were reached. It will also be accompanied by a brief introductory review of the PET physics and PET images properties affecting the segmentation process.

Learning Objectives:
1. Understand the PET image properties affecting the segmentation process
2. Understand the different types of PET-AS algorithms and their advantages and limitations
3. Understand the components and steps of a procedure for evaluating and accepting PET-AS tools and their limitations and dependencies
4. Understand the structure, operation and future prospects of a software tool for evaluation of PET-AS methods


Funding Support, Disclosures, and Conflict of Interest: Funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

Handouts


Contact Email: