Abstract:Background: The extent of skull base invasion (SBI) in nasopharyngeal carcinoma (NPC) directly impacts tumor staging, treatment strategies, and prognosis assessment for NPC patients, emphasizing the critical need for prompt diagnosis and precise assessment of invasion. Thus, we aimed to integrate the advantages of intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) and conventional magnetic resonance imaging (cMRI), and assess their combined diagnostic efficacy versus that of 18F-sodium fluoride (18F-NaF) positron emission tomography/computed tomography (PET/CT) for detecting SBI in NPC patients. Methods: The study prospectively and randomly recruited 62 patients newly diagnosed with NPC by pathological biopsy at the Cancer Center of Affiliated Hospital of Guangdong Medical University from January 2021 to September 2022. All patients underwent baseline cMRI, IVIM-DWI, and PET/CT scans. The IVIM-DWI analysis included 3 primary parameters: true diffusion coefficient (D), pseudodiffusion coefficient (D*), and pseudodiffusion fraction (f). SBI was defined as the involvement of any substructure confirmed by follow-up MRI and clinical symptoms. Inter-observer agreement was evaluated utilizing the intraclass correlation coefficients (ICC) and kappa coefficients. Receiver operating characteristic (ROC) curve was used to evaluate the diagnostic performance of cMRI, IVIM-DWI plus cMRI, and PET/CT. DeLong test was used to compare the areas under the curve (AUC) of the 3 modalities. Results: Excellent inter-observer reliability was observed (range, 0.841-0.946). Among the IVIM-DWI parameters, D* + f demonstrated comparable accuracy to D + D* + f (AUC 0.906 vs. 0.904; sensitivity 88.9% vs. 89.8%; specificity 92.3% vs. 91.0%). IVIM-DWI plus cMRI yielded an overall AUC of 0.947, sensitivity of 92.6%, and specificity of 96.8%, surpassing cMRI alone with an AUC of 0.914 (P=0.025), sensitivity of 91.2%, and specificity of 91.7%, as well as 18F-NaF PET/CT with an AUC of 0.852 (P<0.001), sensitivity of 80.1%, and specificity of 90.4%. In detecting substructures of SBI, IVIM-DWI plus cMRI showed superior performance compared to 18F-NaF PET/CT within the petrous part of the temporal bone (AUC 0.968 vs. 0.871, P=0.011; sensitivity 93.5% vs. 87.1%, specificity 100% vs. 87.1%), pterygopalatine fossa (AUC 0.935 vs. 0.831, P=0.032; sensitivity 93.9% vs. 69.7%, specificity 93.1% vs. 96.6%), and foramen ovale (AUC 0.885 vs. 0.710, P=0.019; sensitivity 76.9% vs. 61.5%, specificity 100% vs. 80.6%). Conclusions: IVIM-DWI plus cMRI can accurately detect SBI and the substructures in NPC, providing a valuable reference for personalized treatment strategies and precise prognosis assessment.

Deep learning model using planar whole-body bone scintigraphy for diagnosis of skull base invasion in patients with nasopharyngeal carcinoma

Diagnosis of skull-base invasion by nasopharyngeal tumors on CT with a deep-learning approach

Deep Neural Network Based Artificial Intelligence Assisted Diagnosis of Bone Scintigraphy for Cancer Bone Metastasis.

Deep Learning for nasopharyngeal Carcinoma Identification Using Both White Light and Narrow-Band Imaging Endoscopy

A Comparative Analysis of CT and MRI in Detecting Skull Base Invasion of Nasopharyngeal Carcinoma

Deep learning for risk prediction in patients with nasopharyngeal carcinoma using multi-parametric MRIs

Advancing precise diagnosis of nasopharyngeal carcinoma through endoscopy-based radiomics analysis

Deep Learning for Fully Automatic MRI-Based Nasopharyngeal Carcinoma Diagnosis

Computer-Aided Pathologic Diagnosis of Nasopharyngeal Carcinoma Based on Deep Learning

Deep convolution neural network model for automatic risk assessment of patients with non-metastatic nasopharyngeal carcinoma

Dual-energy CT for the detection of skull base invasion in nasopharyngeal carcinoma: comparison of simulated single-energy CT and MRI

Automated Diagnosis of Bone Metastasis by Classifying Bone Scintigrams Using a Self-defined Deep Learning Model

Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) combined with conventional MRI for the detection of skull-base invasion in nasopharyngeal carcinoma: comparison with 18F-sodium fluoride (18F-NaF) positron emission tomography/computed tomography (PET/CT)

A Deep Learning MRI-based Signature May Provide Risk-Stratification Strategies for Nasopharyngeal Carcinoma

A Comprehensive Review on Radiomics and Deep Learning for Nasopharyngeal Carcinoma Imaging

Survival rate prediction of nasopharyngeal carcinoma patients based on MRI and gene expression using a deep neural network

Automatic identification of suspicious bone metastatic lesions in bone scintigraphy using convolutional neural network

AI diagnostics in bone oncology for predicting bone metastasis in lung cancer patients using DenseNet-264 deep learning model and radiomics

Automated Diagnosis of Bone Metastasis Based on Multi-View Bone Scans Using Attention-Augmented Deep Neural Networks

A semi-automatic deep learning model based on biparametric MRI scanning strategy to predict bone metastases in newly diagnosed prostate cancer patients

CT diagnosis of skull base involvement in nasopharyngeal carcinoma