PST-Radiomics: a PET/CT lymphoma classification method based on pseudo spatial-temporal radiomic features and...

Area of Science:

• Radiomics
• Medical Imaging
• Artificial Intelligence

Background:

• Existing radiomic methods often overlook intra-tumor metabolic heterogeneity (ITMH), a key factor in tumor subtyping.
• This limitation can hinder the accuracy of lymphoma classification using imaging data.

Purpose of the Study:

• To propose a novel pseudo spatial-temporal radiomic method (PST-Radiomics) for enhanced lymphoma classification.
• To leverage intra-tumor metabolic heterogeneity (ITMH) information from PET/CT scans.

Main Methods:

• Developed a multi-threshold gross tumor volume sequence (GTVS) to exploit ITMH.
• Extracted 1D radiomic features from PET images and GTVS to create a pseudo spatial-temporal feature sequence (PSTFS).
• Reshaped PSTFS into 2D pseudo spatial-temporal feature maps (PSTFM) and utilized a light-weighted pseudo spatial-temporal radiomic network (PSTR-Net) for end-to-end learning.

Main Results:

• PST-Radiomics demonstrated superior performance in a PET/CT lymphoma subtype classification task compared to existing radiomic methods.
• Quantitative experiments confirmed the effectiveness of the proposed method.

Conclusions:

• PST-Radiomics successfully incorporates ITMH for improved lymphoma classification.
• The method's feature map visualization indicates complex feature selection and hierarchical feature extraction, underscoring its superiority.