Piramidal Inc. stepped into the limelight today, announcing that it has filled its war chest with $6 million in seed funding to tackle the intricacies of brainwave analysis using a specially developed foundational artificial intelligence model.
Today’s round was led by Y Combinator, with the participation of several other venture capitalists, including Adverb Ventures, Lionheart Ventures, as well as angel investors such as the founders of Intercom Inc., PlanGrid Inc. and Guilded Inc.
The startup is developing what it terms a “barrier-breaking AI model” that’s designed to help neurologists better assess brainwave data. It hopes that its AI model will speed up and improve the accuracy of brain injury diagnoses and reduce patient waiting times.
Piramidal says it has used cutting-edge machine learning techniques to develop a proprietary AI model that can analyze the results of electroencephalography or EEG tests, which are used by doctors to measure electrical activity in the brain. EEG tests are performed in order to assess the extent of brain injuries and diagnose brain-related diseases. The tests involve mapping the patient’s brainwaves, after which a neurologist will manually inspect the data to try and determine what’s wrong.
But although EEG is often the best method for diagnosing brain problems, it’s a time-consuming and error-prone process, with studies suggesting that neurologists still misinterpret the data in 30% of cases.
Crunching brainwaves
Piramidal co-founder and Chief Executive Officer Dimitris Sakellariou told SiliconANGLE that EEG scans are notoriously difficult to understand because they reveal such a vast number of normal and abnormal patterns of brain activity. “There’s a lot of variability of these patterns across individuals,” he added.
A second challenge, Sakellariou mentioned, is that the EEG scans themselves can often take days to perform. “These things make EEG analysis time-consuming and error-prone, and they require highly specialized medical professionals to interpret them and conduct diagnosis,” he said.
Piramidal, which is named after the “pyramidal neurons” that play a crucial role in higher cognitive functions such as learning, memory and decision-making, wants to improve the accuracy and reliability of EEG tests by using AI to process and analyze the data more rapidly, so it can diagnose patient’s conditions faster and more accurately. Its foundational model is initially focused on epilepsy diagnoses, and clinical tests show it has good potential to reduce error rates.
While large language models such as ChatGPT are trained on text, Piramidal’s AI model is trained specifically on human brainwaves, and this knowledge enables it to understand and detect various facets of brain activity.
Kris Pahuja, co-founder and chief product officer, added that the medical science industry previously found it very difficult to build robust computational methods of analyzing EEG due to the variability found in recordings across different patients and devices. But with the widespread availability of generative AI and increased computational resources now available through the cloud, it has become possible to crack the EEG code.
“We can now develop systems capable of analyzing vast volumes of data and identifying associations across thousands of patients,” he said. “This is a task that was previously unimaginable for smaller machine learning models or humans.”
The startup plans to go beyond epilepsy and diagnose every kind of brain condition, and other use cases might include remote monitoring of brainwave abnormalities, real-time brain health tracking, and drug discovery for neurological conditions.
Life saving properties?
Holger Mueller of Constellation Research Inc. said generative AI technology is much better suited for interpreting EEG scans because it has the ability to crunch data at greater scales and learn patterns much more rapidly.
“The real value though, is that generative AI models like Piramidal’s have the ability to look back at vast amounts of longitudinal data and identify suspicious signs in older EEGs that didn’t previously raise any flags,” Mueller said. “Humans will never have the capacity and willingness to study all of this data, but AI software can, and it will potentially save lives by doing it.”
The funding is important for Piramidal, as budgetary constraints are one of the main challenges that prevent AI from being used more extensively in healthcare. A recent article in Forbes underscores this, noting that the high costs involved in creating AI models for specific uses in healthcare means that many such initiatives never get off the ground.
Pahuja said the funding will ensure the company has the money it needs to continue training its model and bring its first product to market.