The effect of harmonization on the variability of PET radiomic features extracted using various segmentation methods

Seyyed Ali Hosseini,Isaac Shiri,Pardis Ghaffarian,Ghasem Hajianfar,Atlas Haddadi Avval,Milad Seyfi,Stijn Servaes,Pedro Rosa-Neto,Habib Zaidi,Mohammad Reza Ay

DOI: https://doi.org/10.1007/s12149-024-01923-7

2024-04-06

Annals of Nuclear Medicine

Abstract:This study aimed to examine the robustness of positron emission tomography (PET) radiomic features extracted via different segmentation methods before and after ComBat harmonization in patients with non-small cell lung cancer (NSCLC).

radiology, nuclear medicine & medical imaging

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to evaluate the impact of different segmentation methods on the variability of radiomics features in positron emission tomography (PET) images of non - small - cell lung cancer (NSCLC) patients, and to explore the effect of using the ComBat harmonization method to reduce the variability of these features. Specifically, the study aims to:

1. **Evaluate the robustness of PET radiomics features**: The robustness of PET radiomics features extracted by different segmentation methods (including manual delineation, K - means, watershed, fuzzy C - means, region - growing, local active contour, and iterative thresholding with different thresholds) before and after applying the ComBat harmonization.

2. **Analyze the influence of different segmentation methods on radiomics features**: Study the influence of different segmentation methods on the reliability (measured by the intra - class correlation coefficient ICC) of the extracted radiomics features, and explore which feature families (such as first - order statistical features, gray - level run - length matrix GLRLM, gray - level dependence matrix GLDM, etc.) are the most robust under different segmentation methods.

3. **Evaluate the effect of ComBat harmonization**: By applying the ComBat harmonization method, reduce the variability of radiomics features caused by different segmentation methods, and evaluate the impact of this harmonization on the performance of machine - learning classifiers, especially in predicting NSCLC recurrence.

4. **Improve the prediction performance of machine - learning models**: Use support vector machine (SVM) classifiers and linear support vector classifiers (LinearSVC) for feature selection, combined with 5 - fold nested cross - validation, to evaluate the impact of ComBat harmonization on the prediction performance of the model, especially the improvement of the watershed segmentation method.

Through these analyses, the study hopes to provide an effective method to reduce the variability of radiomics features caused by different segmentation methods, thereby improving the accuracy and reliability of NSCLC recurrence prediction based on PET images.