Multimodality CT/MRI Radiomics for Lung Nodule Malignancy Suspiciousness Classification

Abstract

Lung cancer is the most common and deadly form of cancer, and its early diagnosis is decisive to the patient’s survival. Computed Tomography (CT) is the gold-standard imaging modality for lung cancer management, but recent studies have shown the potential of Magnetic Resonance Imaging (MRI) in lung cancer diagnosis and how combining multimodality medical images can yield better outcomes. In this study, we evaluated whether the combination of CT and MRI scans from lung cancer patients can leverage a more precise malignancy suspiciousness classification. For such, we registered paired CT and MRI scans from 47 patients, segmented the nodules in each modality, extracted radiomics features, and performed an experiment with an XGBoost classifier, evaluating models’ performance metrics across 30 trials. The same experiment was performed with four sets of features: 1) CT-only; 2) MRI-only; 3) CT and MRI features; 4) CT/MRI fused images. Our results indicate that the image fusion approach can yield significant AUC performance gains over the single modalities models, with an average AUC of 0.794, but feature concatenation is not an adequate strategy for dealing with multimodality data, as its average AUC of 0.770 didn’t indicate any improvement over the single modalities. Additionally, we observed that MRI, with na average AUC of 0.770, has shown significantly better performance than CT, with 0.754, encouraging further studies in MRI as a lung cancer management image modality. Finally, the analysis on the importance of radiomics features reinforced the relevance of features that reflects on morphological characteristics of a nodule, such as its dimension and roundness, as well as texture features that relate to the intratumoral environment, measuring its complexity and homogeneity.