Post-processing steps improve generalisability and robustness of an MRI-based radiogenomic model for human papillomavirus status prediction in oropharyngeal cancer

Abstract

Objectives: To assess the impact of image post-processing steps on the generalisability of MRI-based radiogenomic models. Using a human papillomavirus (HPV) status in oropharyngeal squamous cell carcinoma (OPSCC) prediction model, this study examines the potential of different post-processing strategies to increase its generalisability across data from different centres and image acquisition protocols.

Materials and methods: Contrast-enhanced T1-weighted MR images of OPSCC patients of two cohorts from different centres, with confirmed HPV status, were manually segmented. After radiomic feature extraction, the HPV prediction model trained on a training set with 91 patients was subsequently tested on two independent cohorts: a test set with 62 patients and an externally derived cohort of 157 patients. The data processing options included: data harmonisation, a process to ensure consistency in data from different centres; exclusion of unstable features across different segmentations and scan protocols; and removal of highly correlated features to reduce redundancy.

Results: The predictive model, trained without post-processing, showed high performance on the test set, with an AUC of 0.79 (95% CI: 0.66-0.90, p < 0.001). However, when tested on the external data, the model performed less well, resulting in an AUC of 0.52 (95% CI: 0.45-0.58, p = 0.334). The model's generalisability substantially improved after performing post-processing steps. The AUC for the test set reached 0.76 (95% CI: 0.63-0.87, p < 0.001), while for the external cohort, the predictive model achieved an AUC of 0.73 (95% CI: 0.64-0.81, p < 0.001).

Conclusions: When applied before model development, post-processing steps can enhance the robustness and generalisability of predictive radiogenomics models.

Key points: Question How do post-processing steps impact the generalisability of MRI-based radiogenomic prediction models? Findings Applying post-processing steps, i.e., data harmonisation, identification of stable radiomic features, and removal of correlated features, before model development can improve model robustness and generalisability. Clinical relevance Post-processing steps in MRI radiogenomic model generation lead to reliable non-invasive diagnostic tools for personalised cancer treatment strategies.

Keywords: Human papillomavirus; Imaging genomics; Machine learning; Magnetic resonance imaging; Radiomics.

Fig. 1

Delineation of primary oropharyngeal carcinoma on contrast-enhanced T1-weighted MRI. This figure demonstrates the segmentation strategy for viable tumour tissue (yellow ROI). Peritumoural inflammation was found with higher intensity surrounding the tumour tissue. Additionally, a necrotic mass was seen with low intensity anterior to the tumour tissue. Both peritumoural inflammation and necrotic mass were excluded by segmentation from viable tumour tissue

Fig. 2

General study design. This study analysed MRI images from two NKI and AUMC oropharyngeal squamous cell carcinoma (OPSCC) cohorts with known HPV status. We first acquired and segmented MR images. Subsequently, radiomic features from both cohorts were extracted. We then performed several post-processing steps, including (1) data harmonisation, (2) identification of stable radiomic features across various segmentations and scan protocols, and (3) the removal of highly correlated radiomic features. The final phase involved modelling and a comprehensive evaluation, focusing on each post-processing step’s impact on the predictive model’s robustness and generalisability

Fig. 3

ROC-AUC curve for performance evaluation of the predictive model with diverse post-processing strategies. This figure illustrates ROC-AUC curves to assess the robustness and generalisability of the predictive model under various post-processing strategies. The evaluated post-processing steps include: a No post-processing, b Data harmonisation, c Stability, d Stability + Data harmonisation, e Correlation removal, f Correlation removal + Data harmonisation, g Correlation removal + Stability, and h Correlation removal + Data harmonisation + Stability