Fuzzy multi-class statistical modeling for efficient total lesion metabolic activity estimation from realistic PET...

Area of Science:

• Nuclear Medicine
• Medical Imaging Analysis
• Computational Biology

Background:

• 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) is standard for therapy follow-up.
• Robust biomarkers are needed for predicting early therapy response.
• Total lesion metabolic activity (TLA) shows prognostic value in longitudinal studies.

Purpose of the Study:

• To develop an efficient marker quantification procedure for PET-derived biomarkers.
• To propose a direct estimation formula for TLA and SUVmean using fuzzy multi-class modeling.
• To evaluate the robustness of the proposed method in realistic scenarios.

Main Methods:

• Utilized fuzzy multi-class modeling with a stochastic expectation maximization (SEM) algorithm.
• Fitted a finite mixture model (FMM) to PET image data.
• Simulated and reconstructed a realistic liver lesion for evaluation.

Main Results:

• Proposed a direct estimation formula for TLA and SUVmean from the statistical model.
• Evaluated results against ground truth knowledge.
• Demonstrated robustness in handling background heterogeneities in simulated liver lesions.

Conclusions:

• The proposed fuzzy multi-class modeling approach provides a robust method for quantifying PET-derived markers.
• This method facilitates accurate estimation of TLA and SUVmean for improved prognostic value.
• The technique is suitable for realistic clinical scenarios involving background noise and heterogeneity.