Report of AAPM TG211: Classification and evaluation strategies of
auto-segmentation approaches for PET
M. Hatt
1
, J. Lee
2
, C. Caldwell
3
, I. El Naqa
4
, C.R. Schmidtlein
5
, E. De Bernardi
6
, W. Lu
7
, U.
Nestle
8
, D. Visvikis
1
, T. Shepherd
9
, S. Das
10
, O. Mawlawi
11
, V. Gregoire
2
, H. Schöder
5
, R. Jeraj
12
,
A. Pugachev, E. Spezi, M. MacManus
13
, X. Geets
2
, H. Zaidi
14
, A.S. Kirov
5*
22
Conclusions: Based on the large number of published PET-AS
algorithms and their relative lack of validation, selecting and
implementing one algorithm among those available is challenging.
There is however accumulating evidence in available comparison
studies that PET-AS algorithms relying on advanced image
paradigms perform better than simple threshold-based
approaches.
The second conclusion of this report is that a standard test
(i.e. a benchmark) dedicated to evaluation of both existing and future
PET-AS algorithms needs to be designed. The first steps in designing
this standard are presented in the second half of the report. The primary
intention of this benchmark is to aid clinicians in evaluating and selecting
PET-AS algorithms for use in clinical practice.