AI in POCUS Quality Assurance: The Pros and Cons

Shawn Sethi, DO, FACEP, FPD-AEMUS
Rachel B. Liu, MD, FACEP, FAEMUS

Disclaimer: There are no commercially ready products that can perform broad quality assurance as of yet.

Artificial intelligence is revolutionizing many aspects of healthcare, and its applications in emergency medicine ultrasound are just beginning. One of the most promising areas is the use of AI machine learning models to assist with quality assurance (QA) of ultrasound images. By evaluating ultrasound images for quality and diagnostic accuracy, AI has the potential to streamline the QA process, and enhance patient care. However, as with any emerging technology, AI-assisted QA comes with both advantages and concerns.¹

Pros:

1. Increased Efficiency in Image Review

One of the biggest advantages of these AI models is their ability to automate tasks and thereby reduce the workload of physicians, particularly for more mundane tasks. Traditional QA requires reviewing numerous images to ensure they meet quality standards. AI models can be trained in both supervised and unsupervised fashions to “pre-screen” images, flagging those that meet certain criteria which require closer and more urgent review.^2,3

2. Standardization and Consistency

Human assessment of ultrasound image quality can be subjective and vary significantly between reviewers. AI models, on the other hand, could apply uniform criteria to all images, reducing interobserver variability. This standardization would be particularly beneficial in a large health system with numerous ultrasound images that are being generated every day.⁴

3. Potential for Expanding Ultrasound Accessibility

AI-assisted QA could support rural or resource-limited settings where emergency physicians with ultrasound expertise may not always be available. These departments could benefit from using a combination of technologies such as tele-guidance and remote QA with the assistance of AI models.

Cons:

1. Integration Challenges and Workflow Disruptions

Despite its potential, incorporating AI into existing QA workflow can be challenging. There is significant variability in software platforms and middleware across emergency departments and hospital systems, which may not be immediately compatible with AI-driven QA systems. Additionally, transitioning to an AI-assisted workflow requires training and adaptation, which may temporarily slow down operations. IT security concerns may also limit additional third party access to AI-driven quality assurance platforms.

2. Need for Large and Diverse Datasets

AI models rely on large datasets to learn and improve their accuracy. If an AI system is trained on a limited dataset that does not include a wide range of patient demographics, sonographer experience, body habituses or pathologies, its performance may be inconsistent or biased. To ensure reliability, AI systems must be validated across diverse clinical settings, which requires significant data collection.⁵

3. Risk of Over-Reliance on AI

While AI can enhance ultrasound QA, it should not replace human oversight. Over-reliance on AI-generated assessments may lead to missed errors, especially if the model encounters an uncommon or unexpected imaging artifact. AI should be used as a decision-support tool rather than a replacement for expert physician review. There are additional ethical and legal considerations that must be considered and require review by appropriate regulatory bodies.

The Future:

If implemented cautiously, AI has the potential to enhance the quality and efficiency of ultrasound QA. For now, AI should be seen as a potential assistant rather than a replacement for human expertise. As technology improves and validation studies expand, AI-assisted ultrasound QA will likely play an increasingly vital role in emergency medicine.

References

1. Dicle O. Artificial intelligence in diagnostic ultrasonography. Diagn Interv Radiol. 2023 Jan 31;29(1):40-45.
2. Drukker L, Noble JA, Papageorghiou AT. Introduction to artificial intelligence in ultrasound imaging in obstetrics and gynecology. Ultrasound Obstet Gynecol. 2020 Oct;56(4):498-505.
3. Blaivas M, Arntfield R, White M. DIY AI, deep learning network development for automated image classification in a point-of-care ultrasound quality assurance program. JACEP Open. 2020 Mar 1;1(2):124-131.
4. Deslandes A, Avery JC, Chen HT, Leonardi M, Knox S, Lo G, O'Hara R, Condous G, Hull ML; Collaborators. Intra- and interobserver agreement of proposed objective transvaginal ultrasound image-quality scoring system for use in artificial intelligence algorithm development. Ultrasound Obstet Gynecol. 2025 Jan 24.
5. Kim YH. Artificial intelligence in medical ultrasonography: driving on an unpaved road. 2021 Jul;40(3):313-317.