In the previous year, investments in AI tools targeted at healthcare were estimated to reach $11 billion, illustrating a strong belief in the potential of artificial intelligence to revolutionize this vital sector.
Numerous startups leveraging AI in the healthcare domain aim to enhance efficiency by automating administrative tasks that support patient care. One such company is Hamburg-based Elea, which is focusing on a relatively underappreciated niche — pathology laboratories that analyze patient specimens for disease diagnoses. Elea’s goal is to implement a voice-driven, AI-powered workflow system designed to improve productivity in labs and extend its benefits to other healthcare departments.
The initial AI offering from Elea aims to transform the workflow of clinicians and lab personnel. This solution effectively replaces outdated information systems and conventional practices (such as using Microsoft Office for report writing), transitioning the workflow to an “AI operating system.” This OS employs advanced features like speech-to-text transcription and other automation tools to significantly reduce the time required to generate diagnoses.
After approximately six months of usage with its initial clients, Elea reports that its system has managed to decrease the report generation time for about half of their cases to just two days.
Step-by-step Automation
Dr. Christoph Schröder, CEO and co-founder of Elea, explains that the existing manual workflows in pathology labs present ample opportunities for productivity enhancements through AI. “We essentially revamp the process, automating all steps… [Doctors] communicate with Elea, the medical technical assistants [MTAs] engage with Elea, indicating what they observe and their desired actions,” he clarifies.
“Elea acts as the agent, performing tasks such as preparing slides and staining, thereby expediting processes significantly,” he adds.
“Rather than augmenting anything, this solution replaces the entire infrastructure,” he states, referring to the cloud-based software designed to supplant legacy laboratory systems that still rely on separate applications for different tasks. The vision for the AI OS is to synergize all functions seamlessly.
The startup enhances its platform by utilizing various Large Language Models (LLMs), fine-tuning them with specialized knowledge relevant to pathology labs. The software incorporates speech-to-text capabilities for transcribing staff voice notes, also employing “text-to-structure,” enabling the system to translate these verbal notes into actionable directives that facilitate the AI agent’s operations.
In addition, Elea intends to develop its proprietary foundational model for slide image analysis, as per Schröder’s vision for expanding diagnostic functionalities. However, the primary focus remains on scaling their initial offering effectively.
Elea’s proposition to pathology labs indicates that tasks that might typically take two to three weeks using traditional methodologies can be accomplished within hours or days. The integrated system aims to streamline operations, reducing the cumbersome back-and-forth associated with manually compiling reports, where human errors can often introduce significant delays.
Lab personnel can access the system through an iPad app, Mac app, or web application, catering to various user preferences.
Founded in early 2024, Elea made its debut with its first lab in October after spending time in stealth mode during 2023. Dr. Schröder, with a background in AI applications for autonomous driving at Bosch, Luminar, and Mercedes, leads the effort.
Co-founder Dr. Sebastian Casu, the startup’s CMO, brings over a decade of clinical experience from intensive care, anesthesiology, and emergency departments, alongside a tenure as a medical director for a major hospital chain.
So far, Elea has established a partnership with a prominent German hospital group (details are yet to be revealed) that manages around 70,000 cases each year. The system has already garnered hundreds of users.
Additional clients are expected to go live “soon,” and Schröder notes that the company is considering international expansion, particularly targeting the U.S. market.
Seed Backing
The startup recently disclosed a €4 million seed round it secured last year, led by Fly Ventures and Giant Ventures, which has been used to strengthen its engineering team and roll out its product to initial labs.
While this amount may seem modest compared to the billions flowing into the sector annually, Schröder contends that AI startups need not deploy vast numbers of engineers or millions in funding to thrive. Instead, it’s about optimizing available resources, particularly in healthcare where a focused, departmental approach can mature specific use cases before branching out.
Nonetheless, he affirms that the team is gearing up for a larger Series A round likely this summer, shifting its efforts to actively market to attract more laboratory clients rather than relying solely on word-of-mouth.
When discussing their strategy in relation to the competitive landscape for AI healthcare solutions, he explains: “I believe the significant distinction lies in our approach as a specific solution instead of a vertically integrated model.”
“Many tools available are merely add-ons to existing systems [like EHR systems] … Users are often required to layer additional tools, which can be cumbersome for those reluctant to engage with digital technology, hindering overall efficiency.”
“In contrast, our solution is deeply integrated into our laboratory information system — or what we term the pathology operating system — meaning users don’t need to navigate separate interfaces or tools. They simply communicate with Elea, conveying observations and instructions,” he elaborates.
“Moreover, there’s no necessity for an extensive engineering team—just a handful of highly skilled engineers will suffice,” he asserts. “With around two dozen engineers in our team, we can achieve remarkable advancements.”
“Today’s most rapidly growing companies do not rely on vast engineering teams; they leverage a small group of specialists to create impressive innovations. This is the philosophy we embrace, which is why we did not initially require hundreds of millions in funding,” he concludes.
“This signals a definitive paradigm shift … in company development strategies.”
Scaling a Workflow Mindset
Opting to focus on pathology laboratories was a tactical decision for Elea, as the addressable market is valued in the billions, according to Schröder. He views the pathology sector as “extremely global,” with the capability for lab companies and suppliers to enhance scalability, especially compared to the more fragmented hospital supply landscape.
“For us, this is incredibly appealing since we can develop one application and scale it from Germany to the U.K. and the U.S.,” he notes. “Because lab workflows are largely similar across these regions, solving this issue in German translates to addressing it in English or Spanish as well, unlocking multiple opportunities.”
He further highlights pathology labs as “one of the fastest expanding fields in medicine.” Advancements in medical science, including the rise of molecular pathology and DNA sequencing, have sparked demand for a broader range of analyses and an increased frequency of testing. This growth results in more responsibilities for labs and intensified demands for productivity.
Once Elea sufficiently refines its lab application, Schröder states that they might explore sectors where AI is more commonly applied in healthcare, such as aiding hospital doctors in recording patient interactions. However, any future applications will maintain a firm emphasis on workflow optimization.
“We intend to instill a workflow mindset, treating every task as part of a workflow culminating in a report that needs distribution,” he explains, mentioning that while they won’t delve into diagnostics within a hospital setting, they aim to “focus on refining operational workflows.”
Image analysis is another promising pursuit for Elea in future healthcare applications, such as expediting data processing in radiology.
Challenges
What about accuracy? Given the sensitive nature of healthcare, inaccuracies in AI transcriptions—such as those related to biopsies checking for cancer—could have profound consequences if discrepancies arise between a physician’s observations and Elea’s transcriptions.
Currently, Schröder mentions they assess accuracy by tracking how often users modify reports generated by the AI. At present, he estimates that adjustments are made in 5% to 10% of cases, which may suggest potential errors. (Though it’s essential to note that physicians might alter records for various other reasons; they are working to proactively reduce the frequency of manual corrections.)
Ultimately, responsibility lies with the physicians and staff who must review and validate the AI-generated outputs, suggesting that Elea’s workflow is not substantially different from the legacy systems it aims to replace (where a doctor’s voice note would also be transcribed by a human, potentially leading to errors). With Elea, “the initial transcription is performed by AI rather than a typist,” he affirms.
Increased automation could lead to heightened throughput volumes, which may strain existing review processes, as human staff may need to handle a significantly larger volume of data and reports than previously.
In this regard, Schröder acknowledges potential risks. However, he points out that they have instituted a “safety net” feature wherein the AI identifies possible issues—prompting the doctor to reassess findings. “We refer to it as a second pair of eyes,” he remarks, explaining that the system evaluates past reports against current physician comments, offering suggestions and insights.
Patient privacy is another significant concern regarding AI solutions that rely on cloud processing (as Elea does), as opposed to keeping data on-site under the lab’s management. Regarding this, Schröder claims they have addressed “data privacy” concerns by decoupling patient identities from diagnostic outcomes, relying essentially on pseudonymization for compliance.
“Every step of the process ensures anonymity — each function addresses a specific task — and we aggregate the data on the doctor’s device when it needs to be reviewed,” he describes. “Essentially, we use temporary pseudo-identifiers throughout our processing stages which are deleted afterward. When the doctor assesses the patient, the identifiers are combined on the device for their reference.”
“We operate with servers in Europe, ensuring adherence to data privacy regulations,” he continues. “Our primary client is a publicly owned hospital chain considered critical infrastructure in Germany. We had to guarantee data security from a compliance standpoint, and they have approved our processes.”
“In the end, we’ve likely exceeded the necessary standards. It’s always better to exercise caution, especially when handling medical data,” he concludes.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
