How a major Twin Cities healthcare provider is pioneering the use of AI for EEG analysis
Artificial intelligence (AI) has become widely recognized for its potential to enhance many aspects of healthcare. However, most neurologists, epileptologists, and other neuro professionals agree that integrating AI with electroencephalography (EEG) analysis holds particular promise for increasing the efficiency of neuro care teams and ultimately improving health outcomes.
At Allina Health Abbott Northwestern Hospital in Minneapolis, Minnesota, which is affiliated with multiple hospitals in the area, Dr. Ram Mohan Sankaraneni, a Board-certified neurologist and Fellow of the American Epilepsy Society and the American Academy of Neurology, has transformed AI’s promise into proof. His goal is to help patients become seizure free so they can have a normal lifestyle. In the summer of 2024, Dr. Sankaraneni and his team at Allina Health Abbott Northwestern Hospital became the first providers in the United States to implement autoSCORE, the only AI model FDA-cleared for routine EEG interpretation.
Developed by Holberg EEG and available exclusively through medtech powerhouse Natus Medical, autoSCORE was trained on the world’s largest dataset of EEG recordings. autoSCORE is the first AI healthcare model capable of automatic and comprehensive clinical EEG analysis.
To grasp the groundbreaking impact of Dr. Sankaraneni’s work with autoSCORE, it’s crucial to first understand the immense challenges facing neuro professionals regarding EEG interpretations. While simple algorithms and spike and seizure detectors have been used in traditional EEG analysis for decades, the practice is still labor-intensive.
Neurologists must scour vast volumes of EEG data, searching for what are often subtle brain activity abnormalities. For patients with brain injuries, epilepsy, and other neurological conditions, reducing the time to diagnose and treat problems can be critical for their health outcomes. Today’s shortages of healthcare workers, particularly the lack of neurologists and skilled EEG technicians in remote and underserved areas, have exasperated this issue.
Researchers, neuroscientists, and physicians have been eyeing AI for the past several years to overcome many of these obstacles. The goal is not to replace human analysis but to use advanced algorithms to process EEG data more rapidly. By streamlining data analysis and automating repetitive and often tedious tasks, neurologists can identify patterns and anomalies more quickly. As a result, clinicians can address complex, critical cases more efficiently, driving better outcomes even when EEG expertise is not widely available.
autoSCORE is an innovative AI solution introduced by Natus and developed by Holberg EEG, designed to tackle many of the challenges of clinical EEG analysis by shortening the time to diagnosis, improving accuracy, and reducing physician workloads. autoSCORE provides expert-level EEG interpretation using deep-learning algorithms trained on more than 30,000 expertly labeled EEG recordings. This enables the AI tool to consistently identify abnormalities, such as epileptiform activity, with accuracy comparable to that of highly trained clinicians.
A 2023 study published in JAMA found that autoSCORE consistently performed with accuracy, sensitivity, and specificity near or above 90%. In addition, researchers found that autoSCORE consistently outperformed other AI models while delivering results on par with leading human experts.
Dr. Sankaraneni’s initial interest in autoSCORE was piqued by his curiosity of where autoSCORE would fit into current workflows
and how it could improve EEG reading, both for him and resident trainees. In addition, he hoped the tool could improve workflow efficiencies by automating some of the required routine analysis for patients with brain disorders and injuries. Dr. Sankaraneni was eager to determine if autoSCORE could more rapidly pinpoint abnormal brain patterns and potential issues that could require investigations.
Driven by his desire to improve the efficiency of his team and confident in the research of his peers, Dr. Sankaraneni’s task now was to see how autoSCORE performed in an actual clinical setting. Could this newly FDA-cleared tool deliver on its promise to reduce the time needed for EEG review?
Dr. Sankaraneni began working with autoSCORE with what he calls a “sense of curiosity.” While introducing autoSCORE required some initial adjustments to his standard workflow, Dr. Sankaraneni noticed an almost immediate positive impact after he began using the tool. In fact, he found autoSCORE’s ability to prioritize EEG studies for review based on the likelihood of abnormalities was remarkable. By flagging the EEG recordings that required the most immediate attention, autoSCORE highlighted the most critical cases rapidly and accurately.
Artifacts, signals that don’t come from the brain itself but are often present in EEG recordings, can obscure underlying issues. Filtering out cardiac, respiratory and other artifacts can be challenging, especially for newer neuro care professionals. Even with studies that included high numbers of artifacts, Dr. Sankaraneni observed that autoSCORE was correctly able to differentiate artifacts from true abnormalities, The tool also helped ensure that no important details were being overlooked.
As Dr. Sankaraneni has continued to use autoSCORE, he has been pleasantly surprised with the tools efficiency across the board in identifying and prioritizing studies as abnormal or normal, especially in the face of complications due to the presence of artifacts.
“autoSCORE was particularly effective in detecting problems amidst artifacts,” Dr. Sankaraneni noted, mentioning a specific case where, despite significant interference from patient movement, the AI tool could still differentiate artifact and correctly identify focal abnormalities. An important change to his workflow was that with autoSCORE Dr. Sankaraneni can now focus on critical cases right away rather than at times needing to wait until the end of the day to review all EEGs.
Dr. Sankaraneni’s experience with autoSCORE aligns with new research on the overall implications of human-machine teaming for he
althcare which goes as far as to say that “human teaming with intelligent machines has become a fundamental characteristic of a successful clinical decision support system.”
When asked about what role autoSCORE can play specifically within a neurology practice, Dr. Sankaraneni calls autoSCORE “…a reliable colleague that can provide a second opinion,”
adding that autoSCORE most definitely offers a “powerful pathway to faster interventions.”
The benefits experienced by Dr. Sankaraneni, including improved workflow efficiency and enhanced diagnostic accuracy, underscore the transformative potential of AI tools in EEG analysis. As AI continues to evolve, its role in neurology is likely to expand, offering new opportunities for improving patient care and outcomes.
However, integrating AI in EEG analysis raises some important considerations about the balance between machine assistance and clinical judgment. While AI tools can significantly enhance efficiency and accuracy, they should be considered a complement to, not a replacement for, human expertise. Research published in a 2022 issue of Nature stated that overcoming the perception that AI will replace expertise is key to the broad adoption of AI for EEG interpretation. Implementing best practices for human-machine teaming is crucial to maintaining the integrity of experienced professional judgement while ensuring that AI tools are being used to their full potential.
autoSCORE’s deployment at Abbott Northwestern Hospital represents a significant step forward in the application of AI technology for neurology. As a pioneer in the use of AI for specific patients, Dr. Sankaraneni’s experience provides valuable insights into the practical benefits of AI in EEG analysis.
Looking ahead, it’s clear the broader adoption of AI in EEG analysis could lead to significant advancements, from increasing efficiencies in individual neurology practices to increasing the accessibility of EEG for remote and underserved areas.
Allina Health Abbott Northwestern Hospital’s pioneering use of AI exemplifies how, when incorporated effectively, AI tools like autoSCORE will dramatically boost efficiency for neurologists, epileptologists, and other neuro care professionals, if not help shape the future of neurological care.
autoSCORE
autoSCORE - the world’s first AI based model that provides comprehensive clinical EEG interpretation with accuracy on par with medical experts.
Related Articles
048290 RevA