SPEQTACLE: An automated generalized fuzzy C-means algorithm for tumor delineation in PET

摘要: Purpose: Accurate tumor delineation in positron emission tomography (PET) images is crucial oncology. Although recent methods achieved good results, there still room for improvement regarding tumors with complex shapes, low signal-to-noise ratio, and high levels of uptake heterogeneity. Methods: The authors developed evaluated an original clustering-based method called spatial quantification tumor—Automatic Lp-norm estimation (SPEQTACLE), based on the fuzzy C-means (FCM) algorithm a generalization exploiting Hilbertian norm to more accurately account non-Gaussian distributions PET images. An automatic reproducible scheme image-by-image basis was developed. Robustness assessed by studying consistency results obtained multiple acquisitions NEMA phantom three different scanners varying acquisition parameters. Accuracy using classification errors (CEs) simulated clinical SPEQTACLE compared another FCM implementation, local information (FLICM) locally adaptive Bayesian (FLAB). Results: demonstrated level robustness similar FLAB (variability 14% ± 9% vs 7%, p = 0.15) higher than FLICM (45% 18%, < 0.0001), improved accuracy lower CE (14% 11%) over bothmore » (29% 29%) (22% 20%) Improvement significant challenging cases 17% 11% 28% 22% (p 0.009) 40% 35% 0.0001). For cases, outperformed (15% 6% 37% 30% 17%, 0.004). Conclusions: benefitted from fully case-by-case basis. This promising approach will be extended multimodal multiclass future developments.« less