Pancreatic Mass Characterization Using IVIM-DKI MRI and Machine Learning-Based Multi-Parametric Texture Analysis

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Artificial Intelligence in Medicine

Background:

• Non-invasive characterization of pancreatic masses is crucial for effective patient management.
• Differentiating between various pancreatic lesions like pancreatic ductal adenocarcinoma (PDAC), pancreatic neuroendocrine tumor (pNET), solid pseudopapillary epithelial neoplasm (SPEN), and mass-forming chronic pancreatitis (MFCP) can be challenging.
• Advanced imaging techniques are needed to improve diagnostic accuracy.

Purpose of the Study:

• To evaluate the efficacy of Intravoxel Incoherent Motion-Diffusion Kurtosis Imaging (IVIM-DKI) combined with machine learning-based texture analysis for differentiating pancreatic mass subtypes.
• To assess the diagnostic performance of IVIM-DKI parameters and texture features in distinguishing between PDAC and other pancreatic lesions.
• To explore the potential of this integrated approach in non-invasive characterization of pancreatic masses.

Main Methods:

• A cohort of 48 biopsy-proven patients with pancreatic masses underwent MRI with IVIM-DKI sequences (14 b-values).
• An IVIM-DKI model with a 3D total variation (TV) penalty function was used to generate parametric maps.
• Texture analysis (TA) was performed on apparent diffusion coefficient (ADC) and IVIM-DKI maps, with features reduced by chi-square test and analyzed using an artificial neural network (ANN) with 5-fold cross-validation.

Main Results:

• Perfusion fraction (f) was significantly higher in pNET compared to PDAC (p < 0.05).
• IVIM-DKI parameters like f and pseudo-diffusion coefficient (D*) showed diagnostic performance (AUC: 0.73-0.77) for differentiating PDAC from MFCP and pNET.
• Texture analysis of PDAC versus non-PDAC using f and combined IVIM-DKI parameters achieved high accuracy (≥ 84.3%) and AUC (≥ 0.84).

Conclusions:

• The IVIM-DKI model incorporating TV texture features demonstrates significant potential for the non-invasive characterization of pancreatic masses.
• This advanced imaging and analysis technique can aid in distinguishing between different types of pancreatic lesions, including PDAC.
• Further validation may establish IVIM-DKI with texture analysis as a valuable tool in the diagnostic pathway for pancreatic masses.