Clinical Scorecard: Integrating Artificial Intelligence into Routine Abdominal Imaging Practices

At a Glance

Key Highlights

• Abdominal imaging accounts for only 4% of commercially available AI radiology applications, much lower than other specialties.
• Current AI tasks focus mainly on automated organ volumetry and detection of systemic liver diseases, with limited emergency imaging AI tools.
• Non-interpretive AI tools improving image quality and workflow are already integrated, but comprehensive diagnostic AI tools for abdominal imaging remain underdeveloped.

Guideline-Based Recommendations

Diagnosis

• AI tools should provide broad diagnostic support capable of flagging acute and suspicious abdominal conditions.
• Radiologists must collaborate with AI experts to train AI systems using large, organ-specific datasets before integration.

Management

• Integrate AI solutions that improve image quality and logistical workflow without disrupting existing PACS systems.
• Adopt AI tools as supportive aids to radiologists rather than replacements to maintain clinical responsibility and trust.

Monitoring & Follow-up

• Continuously evaluate AI tool performance and reproducibility over time due to ongoing learning and evolution of AI algorithms.
• Ensure peer-reviewed evidence supports AI tool efficacy before clinical implementation.

Risks

• Limited transparency of AI decision-making ('black box' issue) raises medico-legal and trust concerns.
• Financial and technical challenges exist in integrating multiple AI tools into routine radiology workflows.
• Narrow task-specific AI applications may not meet the diverse diagnostic needs of abdominal radiologists.

Patient & Prescribing Data

Patients undergoing abdominal CT imaging, including emergency cases with acute abdominal pain

AI tools currently assist in improving image quality and detecting specific conditions but require further development for comprehensive diagnostic support.

Clinical Best Practices

• Use AI to enhance image quality through deep-learning reconstruction methods to reduce noise and artifacts while maintaining low radiation exposure.
• Prioritize development of AI tools that address multiple abdominal organs and diverse pathologies simultaneously.
• Maintain radiologist oversight and responsibility when using AI to ensure diagnostic accuracy and medico-legal clarity.
• Foster collaboration between radiologists, AI developers, and patients to build trust and optimize AI integration.

References

• Expansion of AI tools in radiology and their distribution across specialties
• Meta-analyses on AI application frequency in abdominal imaging
• AI algorithm detecting abdominal free gas, fluid, and fat stranding with high sensitivity and specificity
• Deep learning system for renal stone detection on low-dose CT
• Necessity of thorough development and testing of AI tools before clinical use
• Development of logistical AI tools for imaging protocol selection and quality control
• Deep-learning reconstruction (DLR) algorithms improving CT image quality
• DLR methods enhancing contrast-to-noise ratio and lesion detectability
• DLR improving spatial resolution and diagnostic confidence for hepatic lesions
• Use of thin slices with DLR maintaining low radiation exposure