Technology

Introducing the key technologies that
make Medical AI industry leader.

MAI Machine learning pipeline

The Healthcare AI Machine Learning Pipeline System is designed to foster collaboration between AI researchers, developers, and medical staff, ensuring seamless communication and information sharing. This integrated platform minimizes misunderstandings and promotes real-world, medically-informed AI development. The outcome is faster, reliable AI-driven healthcare solutions that cater to actual patient needs. Efficient communication and collaboration within the system not only ensure high-quality AI products but also lead to significant cost savings.

MAI Machine learning pipeline operates as such:
  1. ① Collect ECGs via AiTiA and public data sources.
  2. ② The collected data undergoes expert review and interpretation to ensure accurate labeling.
  3. ③ AI researchers develop and verify models optimized for servicing and store them in the learning model candidate pool.
  4. ④ The automated learning system is a high-performance model generating system designed
    to utilize all data to tune and enhance the learning model candidates.
  5. ⑤ The system selects the most suitable model and deploys it to an appropriate service center.

ECG Reconstruction:
Lead-to-Lead translation

We possess an ECG generating technology. This allows us to extract representative styles from any given lead and, by combining the intricate styles that appear depending on the angle of measuring the heart, we can generate ECG data for the desired lead. With this technology, we anticipate that 1-lead ECG data collected from wearable devices can be amplified to closely resemble a standard 12-lead ECG, enabling more detailed diagnoses.

ECG Reconstruction operates as such:
  1. ① From our vast pool of ECG data, we have the insight to recognize lead characteristics patterns of particular heart conditions.
  2. ② Therefore, based on a few lead data points, we can infer the heart's conditions and risk factors for disease.
  3. ③ We further extract detailed (positional) signaling data from the heart.
  4. ④ By mapping the detailed (positional/angular) signaling data onto the given leads, we generate 12 leads.

ECG XAI:
Transparent and explainable ECG models

We have Explainable Artificial Intelligence (XAI) technology to interpret our AI ECG models. XAI technology provides a detailed explanation of why and where the ECG model made its decisions. These specific explanations offer trustworthy and clear evidence to all users, especially medical staff. Furthermore, we expect that such outcomes of XAI contirubutes new insights that were previously unknown.

ECG XAI operates as such:
  1. ① Given artificial intelligence ECG model, XAI technology identifies which regions on the ECG had a significant impact to decide model's prediction.
  2. ② Given an abnormal ECG, the generative AI produces a counterfactual normal ECG to compare changes.
  3. ③ Our XAI technology tells us why and where our AI model makes decisions regarding the classification of a given ECG.