Ahead of RSNA 2024, we spoke with Dr. Jacob J. Visser, a radiologist deeply involved in clinical practice and radiology societies such as ESR and RSNA, to discuss the evolution of AI in radiology, the challenges of value-based care, and the critical role of training.

 

From your perspective and experience, what are the primary challenges that radiologists face today in delivering value-based care?

Radiologists face primary challenges in value-based care, including aligning imaging outcomes with patient-centered metrics, integrating with multidisciplinary teams, and using data-driven technologies to optimize cost-efficiency and effectiveness. Additionally, radiologists must adopt AI and data analytics to enhance decision-making, ensuring these tools improve patient outcomes without overwhelming workflows.

 

How is the integration of AI transforming Imaging technology and the practice of radiology at your institution?

AI integration at our institution is progressing, though adoption is slower than expected due to challenges fitting these tools into established workflows. AI has proven valuable in detecting incidental pulmonary embolisms and nodules, enhancing workflow efficiency by automatically identifying potential pathologies. However, fully integrating AI into routine practice remains challenging, especially in complex image interpretation, as aligning AI systems with clinical decision-making and radiologists’ expertise is still evolving.

 

Have you observed significant improvements in workflow efficiency or patient outcomes since implementing AI systems?

We have observed improvements in workflow efficiency with AI, though they are limited to specific cases. For example, AI in bone age assessment significantly reduces analysis time and increases reporting consistency. AI-driven detection of incidental pulmonary embolism (iPE) has also reduced reporting turnaround, allowing faster diagnosis and treatment. However, AI’s impact on efficiency and outcomes is still under investigation for more complex imaging scenarios, with ongoing studies evaluating its effects on workload and workflow integration for radiologists and technologists.

 

How important is standardized data in enhancing AI’s effectiveness in radiology? What challenges do you foresee in achieving widespread data standardization?

Standardized data is essential to enhance AI’s effectiveness in radiology, ensuring models are trained on consistent, high-quality datasets that improve accuracy and generalizability across clinical settings. Standardization helps AI handle variability in imaging protocols, equipment, and reporting formats, leading to consistent results. However, achieving widespread standardization faces challenges, including uniformity across institutions, equipment variations, and data quality levels. Privacy and interoperability issues also complicate data sharing. Addressing these requires industry-wide coordination to develop and implement common data formats, standards, and reporting frameworks.

 

Can you discuss your experience with AI in detecting complex conditions such as lung diseases? How critical is training for radiologists and technologists to use these AI tools effectively?

In my experience with AI in detecting lung nodules, it enhances early detection by identifying nodules that might go unnoticed. AI’s rapid processing of large data volumes to flag suspicious areas has been valuable in streamlining workflows. Beyond detection, AI’s integration into the care pathway is where its potential lies; it assists in monitoring nodule progression over time, supporting decisions on follow-up imaging, biopsy, or further intervention. By incorporating AI at various points—from detection to follow-up—radiologists can make more informed decisions, ultimately improving patient outcomes.

Training for radiologists and technologists is essential for effective AI use. While AI aids detection, radiologists must know how to integrate AI results, interpret insights accurately, and oversee algorithm outputs vigilantly.

 

Have AI technologies helped you identify lesions or anomalies that might have been missed during initial evaluations?

AI technologies have shown potential in identifying lesions or anomalies that might be missed during initial evaluations, especially subtle findings like small lung nodules or early-stage lesions that can be overlooked due to size or ambiguous appearance. In my experience with AI in lung nodule detection, these tools add an extra layer of vigilance, flagging areas needing further attention and enhancing the sensitivity of diagnostic evaluations.

 

Has AI improved your confidence in diagnostic decisions when faced with clinical uncertainty?

In cases of clinical uncertainty, AI has boosted confidence in diagnostic decisions by providing quantitative assessments and large dataset comparisons. For instance, when distinguishing between benign and suspicious nodules is unclear, AI adds valuable information through pattern recognition and volumetric analysis. This additional data supports more informed decisions, especially in borderline cases where clinical judgment alone may leave room for doubt. While AI doesn’t replace radiologist expertise, it augments decision-making by reducing uncertainty and offering consistent, reproducible evaluations across a broader range of parameters.

 

As someone deeply involved in clinical practice and active in radiology societies like the ESR and RSNA, what steps are necessary to prepare the next generation of radiologists and technologists for an AI-integrated healthcare environment?

First, AI education should be integrated into radiology training, helping students understand AI’s capabilities and limitations, including how to evaluate AI outputs and incorporate them into clinical workflows critically. Second, continuous education is essential, with societies like ESR and RSNA providing ongoing training and certifications to keep professionals current with AI advancements. Finally, fostering interdisciplinary collaboration between radiologists, data scientists, and engineers will bridge the gap between technology and clinical practice, ensuring effective AI integration into healthcare systems.

 

With your extensive involvement in shaping the future of medical imaging, what drives your passion for integrating AI into radiology, and how do you envision these efforts transforming patient care in the coming years?

My passion for integrating AI into radiology stems from its potential to enhance diagnostic accuracy and efficiency, ultimately improving patient care. AI can analyze vast amounts of imaging data, identifying subtle patterns and anomalies that may be challenging even for experienced radiologists, supporting faster, more informed decisions, and enabling earlier disease detection, which is crucial for outcomes.

I envision AI transforming patient care by enabling personalized, precision-driven medicine. AI will improve predictive diagnostics, allowing radiologists to provide tailored treatment recommendations based on imaging biomarkers and patient data. By automating routine tasks, AI will also free radiologists to focus on complex cases and patient interactions, ensuring higher quality care. As these technologies evolve, the collaboration between AI and radiologists will become essential to the future of medical imaging, benefiting clinicians and patients alike.