Cardiomatics, a Polish-based AI company in health tech, has raised $3.2M to expand its use of its ECG reading automation technology.
The round will be led by Central and Eastern European VC Kaya. Nina Capital, Nova Capital, and Innovation Nest are also involved.
A $1M non-equity grant is also included in the seed raise from the Polish National Centre of Research and Development.
The startup was founded in 2017 and sells a cloud-based tool that speeds up diagnosis and improves efficiency for clinicians, cardiologists, and other healthcare professionals. It interprets ECGs automatically and can detect and analyse some 20 abnormalities or disorders. The software generates reports on scans in minutes and is faster than any trained human specialist.
Cardiomatics claims its technology is helping to decentralize healthcare access. The tool allows cardiologists to optimize their workflow to see more patients and to treat them faster. It claims that it allows smaller practices and GPs to provide ECG analysis to patients, without the need to refer them to specialist hospital.
According to the startup, more than 700 customers are using its software in more than 10+ countries including Germany, Switzerland, Denmark and Poland. The AI tool has already analyzed over 3 million hours of ECG signals.
It can integrate with more than 25 ECG monitor devices at this point. The software also boasts a modern cloud-based interface that is different from legacy medical software.
When we asked how accurate its AIs ECG readings are, the startup replied that it has validated their accuracy. The data set used to create algorithms includes more than 10 million heartbeats taken from around 100,000 patients. It is continuously growing. We have created the majority of these data-sets ourselves. The rest are available in public databases.
90% of the data is used for training purposes, 10% for validation and testing and 10% to test algorithms. The data-centric AI places great importance on the test sets in order to ensure that they are representative of the signals received from clients. The accuracy of algorithms is checked in experimental work as part of continuous data and algorithm development. We do this with a frequency of one per month. It is checked by our clients every day in clinical practice.
Cardiomatics stated that it will use the seed financing to invest in product development and expand its business activities within existing markets, as well as to prepare for launching into new markets.
The proceeds of the round will be used for fast-paced expansion plans in Europe. This includes scaling up our AI technology to the top market and making sure physicians have the best possible experience. We are preparing the product for launch in new markets. It added that future plans include FDA certification and entry into the US market.
Although the European Medical Device Certification was granted to the AI tool in 2018, it is worth noting that the European Union's regulatory regime for medical equipment and AI is constantly evolving. In May, the EU Medical Device Regulation (now known as the EU Medial Devices Directive) was updated.
The Artificial Intelligence Act, which is a new framework that assesses risk in applications of AI, is also coming. It will likely increase compliance requirements for AI tools such as Cardiomatics. This will introduce requirements like safety, reliability, and lack of bias in the automated results.
It said that when it launched in 2018, it was one of the first AI-based medical devices to be approved in Europe. We monitor the European situation and work closely with the EU to create a framework that is risk-based for AI regulation. This allows us to keep up with the times. We monitor any draft regulations or requirements that might be proposed. We will implement any new standards or requirements for artificial intelligence in the company and product operations.
It also acknowledged that it is difficult to objectively measure the efficacy ECG reading algorithms.
TechCrunch said that it can be difficult to assess the effectiveness of algorithms objectively. It is usually performed using a small set of data, from one patient. We can receive signals from different patients from different recorders. This is a method we are currently working on to assess effectiveness. They will be able to use our algorithms to accurately evaluate their performance, regardless of the various factors that are part of the study such as the recording device or the social groups being tested.
ECG interpretation by doctors is based on experience, rules, and art. A human interprets an ECG and sees a curve. It operates on a visual layer. An algorithm can see a stream or numbers and not a picture. The task becomes a mathematical problem. It said that you can't build efficient algorithms without knowing the domain. These knowledge and the expertise of our medical team make Cardiomatics a great art form. It is important to remember that algorithms can also be trained from data generated by cardiologists. Machine learning and medical experience are strongly linked.
