The application of AI models has become ubiquitous across various domains, including medicine. However, their inconsistent outcomes raise concerns about their reliability and effectiveness in real-world scenarios. One startup, Piramidal, is working to address this issue with a foundational model for analyzing brain scan data.
The Challenge of EEG Technology
Co-founders Dimitris Sakellariou and Kris Pahuja have observed that electroencephalography (EEG) technology is widely used in hospitals but suffers from fragmentation among different types of machines. This makes it challenging to interpret the data, requiring specialized knowledge and expertise.
"In the neural ICU, there are nurses actually monitoring the patient and looking for signs on the EEG. But sometimes they have to leave the room, and these are acute conditions," said Pahuja. "An abnormal reading or alarm could mean an epileptic episode, a stroke, or even death."
The inconsistent outcomes of AI models in medical applications can be attributed to various factors, including limited training data, biased algorithms, and inadequate model evaluation.
Piramidal’s Approach
Piramidal is working on developing a foundational model for analyzing brain scan data using EEG technology. This model aims to provide consistent and reliable outcomes by addressing the fragmentation issue and improving data interpretation.
"Our goal is to create a warm start for EEG analysis, which can improve the accuracy of diagnoses and treatment plans," said Sakellariou.
The Importance of Data
To develop accurate and effective AI models, high-quality training data is essential. Piramidal has aggregated and harmonized open-source data from various sources to train their foundational model.
"We have enough data to get our first production model trained, but we need more to improve its accuracy," said Pahuja.
The partnerships with hospitals will provide valuable and voluminous training data, which can help elevate the next version of the model beyond human capability.
Moving Forward
Piramidal needs two essential components to move forward: money and data. They have secured a $6 million seed round co-led by Adverb Ventures and Lionheart Ventures, with participation from Y Combinator and angel investors. This funding will be used for compute costs and staffing up.
"To develop accurate AI models, you need access to large amounts of high-quality training data," said Sakellariou. "We’re working on aggregating and harmonizing existing open-source data and collecting more through our hospital partnerships."
Conclusion
The inconsistent outcomes of AI models in medical applications highlight the need for more reliable and effective solutions. Piramidal’s foundational model for analyzing brain scan data using EEG technology addresses some of these challenges by providing a warm start and improving data interpretation.
While superhuman capability is not necessary to improve the quality of care, rigorous evaluation and documentation are crucial. The ICU pilots will allow the tech to be evaluated and documented much more rigorously, both in scientific literature and likely in investors’ meeting rooms.
Key Takeaways
- AI models for medical applications require high-quality training data.
- Piramidal’s foundational model addresses the fragmentation issue and improves data interpretation using EEG technology.
- Partnerships with hospitals will provide valuable and voluminous training data.
- The ICU pilots will allow the tech to be evaluated and documented rigorously.
Related Topics
- AI
- Biotech & Health
- eeg
- Exclusive
- Piramidal
- Startups
