AI Cardiologist
Real-time heart monitoring and alerting dangerous arrhythmia by an AI-based detection and classification module based on RoBERTa LLM
Real-time heart monitoring and alerting dangerous arrhythmia by an AI-based detection and classification module based on RoBERTa LLM
Improve the AI-based arrhythmia detection that classifies dangerous arrhythmia even if the patient is physically active and the ECG is corrupted.
1.Although wearable technology offers freedom of movement, still it is a source of a lot of noise that corrupts the measured ECG and in the case of physical activity, the ECG data may become uninterpretable.
2. Wrong detection and classification of arrhythmia may introduce a lot of false alerts, thus, it can not be a reliable machine-based automated medical device system without clarification by the doctor.
1.Use a transformer-based language model that uses self-attention to process input data sequences and generate contextualized representations of ECG samples, heartbeats, and their relative location.
2.Google's BERT as a deep learning model to be used developing tokens from ECG features, building a heart-based language, and applying transformers to analyze sequences of time-series ECG data to detect and classify dangerous arrhythmia.
1.Bring an AI doctor closer to the end-user, alerting dangerous arrhythmia and therefore, preventing severe heart damage, with results understandable by patients.
2.Patients with diagnosed heart problems, people at risk of high blood pressure, and all elderly cannot monitor their hearts unless they are hospitalized.
Different training/testing ECG datasets,
Improved F1 score from 90% to 95%,
Reduced the error rate by double (from 10% to 5%).
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 951847