VC funds for AI tools for medical care were expected to record $ 11 billion last year. This is the headline level that speaks of an extensive guilty ruling that artificial intelligence is an important sector.
Many startups that apply AI to medical care are trying to promote efficiency by automating some of the management that turns off orbits and enables patient treatment. Hamburg -based ELEA is greatly suitable for this mold, but the pathology laboratory (pathological laboratory) begins with a relatively overlooked and marginalized niche market to analyze the patient sample for the disease. It also includes implanting a workflow -centered approach to accelerate the production of other medical departments.
ELEA’s initial AI tools are designed to check the operation of clinical and other laboratory staff. It completely replaces the legacy information system and other specified methods (for example, using a Microsoft office to enter a report on the use of Microsoft Office). Move the workflow to the “AI operating system” so that the time it takes to output the diagnostic A diagnostic A and other automation is to “substantially”.
ELEA said that it has been operated with the first user for about a year, and the system can reduce the time to lower half of the report in two days in the laboratory.
Step -by -step automation
Dr. Christoph Schröder, the CEO and co -founder of ELEA, often says that it often means manual workflow, which can increase productivity by increasing productivity by applying AI. “We basically overturn all these steps. All steps are much more automated … (doctors) talk with ELEA, the Medical Technical Assistant (MTA) talks with ELEA, and tell me what they want to see.”
“ELEA is an agent and performs all the tasks of the system and prints things. For example, preparing slides such as stains and everything to work much faster, much faster, much more smoother.”
“In fact, we replace the entire infrastructure without strengthening anything.” He adds cloud -based software that wants to perform various tasks using individual apps by replacing the legacy system of the laboratory and more work. The idea for AI OS is that you can adjust everything.
Startups are building a variety of large language models (LLMs) that are fine -tuned with professional information and data to enable core functions in the pathology laboratory context. The platform is baked in speech-text, and the employees’ voice notes and “text-structure” are also killed. That is, the system may convert these transcriptional voice notes in the active direction of supplying power to the operation of the AI agent, which may include sending guidance to the laboratory kit to keep the workflow.
ELEA also plans to develop its own basic model for analysis of slide image per schröder for developing diagnostic functions. However, we are currently focusing on expanding the initial offering.
The pitch for the startup’s laboratory suggests that the integrated system can take two to three weeks within several hours or days, and that the integrated system can be achieved within a few hours or days by replacing things that surround the manual and others of the report and surrounding the terrain that can be mainstream of many frictions.
This system provides a variety of touchpoints that can be accessible to the laboratory staff through the iPad, Mac app or web app, and are suitable for various types of users.
The project was founded in early 2024 and spent the stealth working time in 2023, with the background of applying AI to the autonomous driving projects of BOSCH, Luminar and Mercedes with the first laboratory in October.
Another co -founder, SEBASTIAN CASU, Dr. Startup’s CMO, has brought clinical backgrounds and has been a medical officer of a large hospital chain, working in intensive care, anesthesia and emergency rooms for more than 10 years.
So far, ELEA has partnership with major German hospital groups (not yet disclosed). Therefore, the system has hundreds of users so far.
More customers will be released “soon” and Schröder also says that it is also seeing an international expansion that catches special attention to entering the US market.
Seeding
Startups are unveiling 4 million euros for the first time last year, and the led by Fly Ventures and Giant Ventures were used to organize the engineering team and get the product in the first laboratory.
This figure is quite small, with billions of funds mentioned earlier and now every year. But Schröder claims that AI new companies are not successful of engineers and hundreds of millions of people. And in this health care context, this means taking a department -oriented approach and maturing the target use case before going to the next application area.
Nevertheless, at the same time, he said he would try to post the A round of the series this summer.
He discusses the competitive environment of the AI solution in the medical field and their approach, and tells us as follows.
“Many tools you see are add -ons of existing systems (eg EHR systems).
“Instead, what we have built is actually called our own laboratory information system or called pathology operating system. This means that ultimately, users don’t even need to use other UI and do not need to use other tools. And it just talks with ELEA, tells what I want to see, and tells Elea’s work to do in the system. ”
“You no longer need a lot of engineers. You need 12, really 24 people, really good things,” he claims. “We have approximately 24 engineers on the team, and they can do something amazing.”
“There are no hundreds of engineers in the fastest growing company these days. There are one or two experts, and they can make amazing things. This is also the philosophy we have. So at least we don’t have to raise hundreds of millions of times, ”he adds.
“It is clearly a paradigm shift in the way you build a company.”
Workflow Mind Scaling
The choice to start as a pathological laboratory was an ELEA’s strategic choice, not only worth billions of dollars of dollars in the Schröder, as well as billions of dollars in the world, as well as the expansion of the software by softening the pathological space, “very worldwide”.
“It’s very interesting because we can build a single application and actually expand from Germany to the United Kingdom and the United States.” “Everyone thinks the same, acts the same, and has the same workflow. And if you solve it in German, the great point of LLM is also solved in English (and other languages such as Spanish), so a variety of opportunities are held. ”
He also praises the pathology laboratory as “one of the fastest growing areas in medicine.” This points out that the development of medicine, such as molecular pathology and increased DNA sequencing, creates more types of analysis and demand for more analysis. All of this means more tasks about the laboratory and puts pressure that the laboratory is more productive.
After ELEA has matured the use of laboratory use, the AI can move to an area that is more commonly applied to medical care, such as supporting a hospital doctor to capture patient interactions, but other applications will focus on workflows.
“What we want to bring is that the way of thinking of this workflow is a place where everything is handled like a workflow work, and there is a report, and a report must be sent.”
Image processing is another area that is interested in other future medical applications, such as ELEA, such as increasing data analysis with radiation.
challenge
How is the accuracy? Health care is a very sensitive use case, so it can result in serious consequences if there is a disagreement between the human doctor’s speaking and reporting to other decision makers in the patient therapeutic chain in connection with the error of these AI warriors.
Schröder now says that the AI’s report evaluates the accuracy by looking at how many users change the character. Currently, he says that there are 5% to 10% of these automatic reports that can indicate errors and some manual interactions. (He also suggests that a doctor may have to change for other reasons, but he says he is trying to “promote” the rate of manual intervention.)
Ultimately, Buck is stopping with doctors and other employees who are asked to review and approve AI output. It suggests that the workflow of the ELEA is not different from the legacy process designed to be replaced (eg, the voice of the doctor can be included in an error, a Typist ”).
Automation can increase throughput volume volume, which can apply the checks that human employees need to process more data and reports than before.
As a result, Schröder agrees that there may be danger. But he says that AI has built a “safety net” feature to discover potential problems. Use the prompt to encourage the doctor to see it again. “We call this the second eye,” he added: “We evaluate the previous results report to the doctor right now and where to provide opinions and suggestions. “
Patient confidentiality can be another concern attached to agent AI that depends on cloud-based processing, not data under the control of on-premises and laboratory. As a result, Schröder claims that a new company has solved the “data privacy” problem by separating the patient’s identity from the diagnostic output. Therefore, basically depends on the name of the data protection regulations.
“All steps are always anonymous to do one thing. We combine the data of the device that the doctor sees them,” he says. “Therefore, we basically use doctors IDs used in all processing stages (used in all processing stages that are deleted later, later, but they are coupled to the device for him during the time when the doctor looks at the patient.”
“We cooperate with the European server to see if everything is complying with data personal information,” he says. “Our lead customer is a publicly owned hospital chain called Critical Infrastructure in Germany. From the perspective of the data protection of data, we had to confirm that everything was safe. And they gave us a thumb. ”
“Ultimately, we probably have been excessively expanded. However, especially when treating medical data, it is always better to be on a safe side. ”
