The Biden administration’s recent executive order on artificial intelligence (AI) has significant implications for radiology, as discussed in a review published in JACR. The order aims to ensure responsible use of AI in healthcare and establish a federal program to address unsafe practices. While immediate changes to radiology practice may not be expected, the order signals forthcoming regulatory shifts, particularly in oversight and enforcement by government agencies. This includes scrutiny of computer-aided detection systems and AI for noninterpretative tasks. The FDA premarket review for medical devices like CAD programs is likely to be augmented with additional quality and equity requirements. Health and Human Services will oversee data input into AI algorithms, possibly mandating disclosure of training datasets. Radiologists seeking Medicare reimbursement for AI products will need to prioritize security and compliance with nondiscrimination laws. The order also emphasizes data sharing with the National AI Research Resource and encourages radiologists to engage in policy creation and provide input on regulatory frameworks. However, specific parameters and regulatory details are yet to be defined. Overall, the order serves as a call for federal agencies to mobilize efforts in AI oversight, with radiologists urged to actively participate in shaping policies and best practices.
Predicting Lymph Node Metastasis
A recent study published in Radiology: Imaging Cancer compared the effectiveness of a four-dimensional (4D) convolutional neural network (CNN) model, incorporating clinical and breast MRI findings, with a machine learning model based solely on clinicopathologic features in predicting axillary node status in women with breast cancer. The 4D CNN model achieved a significantly higher area under the curve (AUC) of 87% compared to 63% for the clinical model. It also demonstrated higher sensitivity (89% vs. 75%) and specificity (76% vs. 52%), with a lower false-negative rate (11% vs. 25%). The study suggests that the 4D hybrid model could serve as a valuable tool in selecting patients who may avoid invasive procedures like sentinel lymph node biopsy and aid in treatment decisions for breast cancer patients. However, further external validation of the model is needed, and limitations such as reliance on manual tumor bounding boxes and specific MRI device imaging need to be addressed for broader clinical adoption.
FDA Clearance for Cardiac and Lung AI
Exo, a medical imaging software and device company, has announced FDA clearance for its cardiac and lung artificial intelligence (AI) applications on Exo Iris, their handheld ultrasound device. This expands Exo’s cleared applications to include cardiac, lung, bladder, hip, and thyroid assessments. Iris, powered by AI, facilitates point-of-care ultrasound, particularly benefiting rural or underserved communities, enabling faster diagnosis and treatment. Sandeep Akkaraju, Exo’s CEO, emphasizes the aim of democratizing AI-empowered medical imaging for all caregivers. The AI applications were trained on a diverse dataset and validated across various patient populations and scan types. They enable reliable assessment of pulmonary edema and cardiac function, with additional doppler capabilities for cardiac, abdominal, and vascular applications. Clinicians, including those with limited experience, welcome the efficiency and reliability of Exo’s AI applications, which enhance patient care and healthcare system efficiency.
Sources:
Radiologybusiness.com
mddionline.com
Openai.com
