AI models are becoming increasingly prevalent in various industries, including the medical field. However, the outcomes of these models are often inconsistent. This inconsistency is particularly noticeable in the field of neurology, where analyzing brain scan data is crucial for diagnosing and treating brain disorders. Piramidal, a startup founded by Dimitris Sakellariou and Kris Pahuja, aims to address this issue by developing a foundational model for analyzing EEG data.
The Need for Consistent and Automated Analysis
The founders of Piramidal recognized that the current EEG technology, although widely used in hospitals, is fragmented and requires specialized knowledge for interpretation. Nurses and doctors often have to manually monitor EEG readings, which can be time-consuming and prone to human error. By developing a software solution that can consistently flag worrisome patterns in EEG data, Piramidal aims to improve outcomes for patients with brain disorders while relieving healthcare professionals of some of their workload.
One of the major challenges in automating the analysis of EEG data is the variability in EEG systems, hospital IT setups, and data formats. Each EEG machine may have a different number of electrodes and electrode placements, making it difficult to create a universal model for interpreting brainwave patterns. Piramidal’s founders believe that a foundational model for EEG readings could streamline the process and make life-saving brainwave pattern detection more accessible.
Building a Foundational Model for Neurology
Piramidal’s foundational model for EEG readings aims to provide a standardized approach to analyzing brain activity. While the model is not intended to be a one-size-fits-all solution, it is designed to be adaptable to different setups, electrode configurations, and patient types. The founders emphasize that their focus has always been on building a practical solution for real-world healthcare settings, rather than purely academic research.
The first production version of Piramidal’s model is set to be deployed in hospitals early next year, with four pilot programs scheduled to test its effectiveness in ICU settings. The company’s goal is not just to provide a foundational model but also to fine-tune it for specific applications. Unlike other AI companies, Piramidal does not plan to monetize its model through API fees but rather aims to add value through its comprehensive and scalable approach to EEG data analysis.
Challenges and Opportunities in AI Development
To continue developing their AI model, Piramidal faces two key challenges: funding and data. The company recently secured a $6 million seed round to cover compute costs and expand its team. Additionally, Piramidal is actively working on aggregating and harmonizing open-source data to enhance the training of its model. Partnerships with hospitals will also provide valuable training data, potentially enabling the model to surpass human capabilities in detecting subtle patterns in EEG readings.
The Path to Enhanced Patient Care
While achieving superhuman capability in AI analysis may be a long-term goal, Piramidal recognizes that even incremental improvements in EEG data interpretation can have a significant impact on patient care. By rigorously evaluating their model through ICU pilots and scientific documentation, Piramidal aims to demonstrate the value of AI technology in neurology and pave the way for more efficient and accurate diagnoses of brain disorders.