Pancreatic ductal adenocarcinoma (PDAC) is the seventh leading cause of cancer mortality worldwide. Radiology plays a central role in its initial diagnosis and staging, determining whether a patient can undergo potentially curative surgery or needs chemotherapy first.
CT is the preferred imaging modality over MRI due to wider availability, better image quality consistency, and lower cost. CT and MRI provide similar sensitivity for tumor detection (76-96% and 83-94% respectively), but CT dominates clinical practice. MRI and PET/CT are reserved for specific problem-solving scenarios.
PDAC staging hinges on whether the tumor involves nearby blood vessels — specifically the superior mesenteric artery, celiac artery, portal vein, and superior mesenteric vein. Based on the degree of tumor contact with these vessels, cancers are classified as resectable, borderline resectable, locally advanced, or metastatic.
Patients with resectable disease can go directly to surgery, while those with borderline or locally advanced disease receive neoadjuvant chemotherapy — sometimes followed by surgery if the tumor shrinks sufficiently. Accurate staging is therefore critical: understaging sends an inoperable patient to surgery, while overstaging denies an operable patient their best chance at cure.
Traditional CT images are displayed as flat, two-dimensional cross-sections. Cinematic rendering is an advanced 3D visualization technique that applies physically accurate lighting and shadow effects, creating photorealistic images that more closely resemble what surgeons see in the operating room.
For PDAC, cinematic rendering can accentuate the subtle texture differences between tumor tissue and normal pancreatic tissue, improving tumor conspicuity — especially for small lesions that may be missed on standard imaging. It also improves visualization of the spatial relationship between the tumor and adjacent blood vessels critical for staging.
AI has demonstrated impressive sensitivity for detecting both solid and cystic pancreatic masses in CT scans — achieving 98-100% sensitivity for solid masses and 92-93% for cysts 1 cm or larger. Radiomics signatures extracted from pre-diagnostic CT scans could identify PDAC with 95.5% sensitivity and 90.3% specificity up to 386 days before formal diagnosis.
For staging accuracy, AI has also shown promise in predicting vascular margin involvement — a key factor in determining surgical resectability. One radiomics model achieved 64.8% sensitivity for predicting margin positivity after resection, significantly outperforming standard NCCN criteria assessment (38.9% sensitivity).
Three areas present persistent challenges for radiologists staging PDAC. First, predicting lymph node metastases is unreliable on CT because size alone is not a sufficient indicator of malignant involvement. AI models analyzing lymph node texture and shape may improve on current radiologist performance.
Second, detecting subtle liver and peritoneal metastases — which would make a patient inoperable — requires meticulous review. Third, assessing response to neoadjuvant chemotherapy is hampered by treatment-induced fibrosis that can mimic residual tumor on imaging, making it difficult to determine if surgery should proceed.
The review concludes that AI has the potential to function as a second reader alongside radiologists, improving detection of small early-stage tumors that might otherwise be missed. Earlier detection shifts more patients toward surgical resection — the only curative option — and could meaningfully improve survival outcomes.
AI also shows promise as a source of imaging biomarkers that predict disease recurrence and patient survival after surgery, which could guide which patients need more aggressive adjuvant therapy. Ongoing research and prospective validation are needed before these tools enter routine clinical practice.