Project / product name: ECGents - Making Ambulance Care Efficient
Link to the project: https://github.com/TndmElephant/ECGents
Team leader: Balazs Molnar
Challenge: 2. EKG-Based Hospitalization Prediction
Problem: Ambulance crew arrives to answer an emergency call by patient with cardiovascular symptoms. In many cases, the patient is taken to the hospital or a specialist clinic, but this course of action may not be necessary in all cases. And indeed, retrospective data from IKEM shows that in 60% of cases, the patient could have been treated as an outpatient. This calls for a system which can suggest where the paramedics should take the patient, contributing to the general efficiency of ambulance care.
Solution: Our solution suggests the best course of action in a cardiac emergency situation that support the decision making of the ambulance crew on site with the patient. The solution requires a short ECG recording of the patient taken by the paramedics crew. An algorithm transforms the signal via continous wavelet transform to an image, and a deep convolutional neural network infers whether the patient should be taken either to a hospital, a cardiology center, or can be treated as an outpatient.
Impact: Our system can have a huge impact on ambulance care, supporting the decision making of the crew on site. The system has the ability to identify patients that need hospitalization and even suggests that the patient should be takent to a specialist clinic. This is done while also reducing the number of times hospitalization is not needed, which is the majority of cases. This can have a large impact on the efficiency of ambulance care, saving time for both the ambulance crew and the clinicians.
Feasibility & financials: The proposed solution is easy to implement in a real-life-scenario. The input ECG data is picked up on site, and there is no other data source that the system needs to make the decision. The system can be integrated on a small computing device, even a mobile phone, if it can be connected to the ECG recorder. The additional financial burden should be small compared to the money and time saving that the system yields on the large scale.
What is new about your solution?: The solution uses a novel idea, that is the transformation of ECG signals to wavelet power spectrum image. The system developed as a whole is again a novel implementation, there are no current readily applicable solutions that support the decision of the ambulance crew in terms of hospitalization of the patient.
What you have built at the hackathon - text explanation + code (e.g. GitHub link): https://github.com/TndmElephant/ECGents This is a PoC for quick asessment of whether a patient needs medical care or even specialized cardiological treatment. The program only requires short (~10 beat) ECG signals, as such it can aid decision making in the earliest stages of diagnosis (e.g. ambulance car), alleviating the need for a cardiologist for every case. Without serious optimizitaion this method achieves up to 89.5% precision for predictions where no further medical care is required.
What you had before the hackathon, please mention open source as well: This is a proof of concept algorithm for quick asessment of whether a patient needs medical care or even specialized cardiological treatment. The program only requires short (~10 beat) ECG signals, as such it can aid decision making in the earliest stages of diagnosis (e.g. ambulance car), alleviating the need for a cardiologist for every case. Without serious optimizitaion this method achieves up to 89.5% precision for predictions where no further medical care is required.
What comes next and what you wish to achieve: We wish to work with clinicians at IKEM to improve the algorithm and possibly bring the solution to market.