Applying Artificial Intelligence Technologies to Detect Pancreatic Ductal Adenocarcinoma Using Routine Laboratory...

Area of Science:

• Oncology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Pancreatic ductal adenocarcinoma (PDAC) is a challenging cancer to detect early.
• Routine laboratory tests (RLTs) are widely available and used in regular health checkups.
• Developing non-invasive methods for early PDAC detection is crucial for improving patient outcomes.

Purpose of the Study:

• To investigate the potential of artificial intelligence (AI) to identify PDAC in patients under 70 years old.
• To evaluate the efficacy of using only routine laboratory tests (RLTs) for PDAC detection.
• To assess the diagnostic performance of self-organizing maps (SOM) and Bayesian regularized neural networks (BRNN) for PDAC identification.

Main Methods:

• Utilized 27 routine laboratory tests (RLTs), including blood and urine analyses and body mass index.
• Employed a self-organizing map (SOM) for data clustering and visualization.
• Applied a Bayesian regularized neural network (BRNN) for diagnostic classification.
• Conducted two independent studies comparing PDAC patients with healthy individuals and PDAC-negative outpatients.

Main Results:

• In Study-I, SOM effectively separated PDAC patients from healthy controls.
• BRNN analysis in Study-I achieved high diagnostic performance (e.g., 96.2% sensitivity, 100% specificity in males).
• Study-II showed promising sensitivity and specificity in both SOM projection and BRNN analysis, though with some variation between genders and patient groups.

Conclusions:

• Both SOM and BRNN demonstrated potential in identifying PDAC patients under 70 using RLTs.
• This pilot study suggests AI analysis of RLTs could be a valuable tool for early PDAC detection.
• Further research with larger cohorts is warranted to validate these findings and refine the AI models.