Starting from the WHO document about needs, requirements and challenge we decide to develop our solution starting from this need:

"Optimize current delivery platforms and develop alternative delivery platforms for essential health services: develop remote work solutions, boost home hospitalization programmes and rapidly scale up existing e-Health strategies"
- WHO -


we observed 3 main problems:

  • detect patients at risk
  • screening the whole population
  • monitor them

People with chronic pulmonary diseases have a higher risk of being infected by covid-19 and have a more negative trend. It is crucial to make a quick diagnosis and monitor these patients.

What it does

With our team competence we create:

  • Real-time analysis & AI elaboration of heart, respiratory diseases and vital parameters (Single-derivative-ekg, Oxygen saturation, Heart rate, Respiratory rate, Blood pressure).
    For the vital parameters we did a long search about the best device for us, we don’t provide hardware WE USE EXISTING HARDWARE, reducing the timing for ideation, prototype and create the chain for production. The real improvement is our knowledge on Software and machine learning MAKING ABLE EVERY COMPATIBLE DEVICE TO WORK WITH OUR SOFTWARE and scalable up to million of users right in a day.

  • Pre-triage online, home remote monitoring, thanks to a test that evaluates symptoms, vital data, risks factor and processes them through a score to classify patients based on their stable, unstable or critical condition

  • Sharing and visualization of scores, parameters and AI elaboration with physician and Hospitals. The anonymous database is very flexible and make use of modern no-sql technologies for securely storing personal data anonymously and share it with physicians and integrate it with hospital and all third parties. Our technology is really flexible and easy to manage letting us able in few days to satisfy every request comes from this kind of organization.

How we built it

  • Scientifically proven Machine Learning and AI algorithms:
    Here we attach the 2017 position paper.

    Essentially the technology used is that presented in that paper. From a technical point of view, the deep learning neural network used in that work was replaced with the newer ResNet-152 giving a leap of accuracy of 93% in F1-Weighted score for arrhythmias classification and 95% F1-Weighted in atrial fibrillations (not yet described in that paper).
    With IntelliHearts we mainly use sensors related to ppg and ECG, and we are able to detect:

    Bradycardia, tachycardia, fusion beats and ventricular and supraventricular ectopic beats.

    In the experiment that was conducted we first acquired the signal and cleaned it with statistical techniques (e.g. wavelet) then inter-patient separation was made. Subsequently every single beat is plotted and shown in 2d centered in the peak R.
    Consequently we have n beats that can fall into the categories: normal, ectopic, supraventricular and fusion. There are not always the same number of beats within the same class, for this reason we refer to the F1 microscore average.
    This is the basis of the paper. From this moment on, the classifier has been changed, as each plotted beat is analyzed by a classifier, which in the paper was a Convnet, but which we have changed and used a ResNet 152 layers. Reaching from 90% of the ConvNet proposed in the paper to 93% of the current ResNet.

    Respiratory diseases detection was conducted in the following way:

    • We used the data from Paper: Α Respiratory Sound Database for the Development of Automated Classification
    • Filter noise and remove background sounds
    • We used several segmentation techniques based on spectrograms of respiratory audio
    • We used a custom deep learning neural network to predict binary healthy/unhealthy patients using only audio recorded by smartphone
    • Developed a web service and web page/app to record and show the result.

  • We use medical algorithms already used in the wards of infectious diseases or pre-surgery to investigate the general state of the patient, taking into account oxygen saturation, temperature, heart rate, respiratory rate, neurological condition. Thanks to the test we investigate the symptoms, the risk pathologies and we can give indications to the patient about what to do

  • Shared databases that could be used by physician for evaluation, motorize and have history of the patient.

Challenges we ran into

VAE/GAN needs lots of data: the majority of reviewed papers use 70-30 split without using the official test set provided by the challenge. Thus they don't use a inter-patient separation scheme, revealing wrong results. Find valid datasets, identify the right devices to work with, competition and hackathon dismissed for covid-19.

What we have done during the weekend

During the weekend we have developed the respiratory disease detection with deep learning algorithm, we develop a demo web app and created a temporary web service for breath audio analysis.
We hit the 95% of accuracy to dectect respiratory diseases.

You can test it at do it from your smartphone and follow the instruction on website.

Accomplishments that we're proud

The segmentation works!! For the classification at moment I have 95% accuracy in binary classification (healthy vs pathology) with interpatient separation scheme testing on the official test set provided originally from the paper.

Accomplishments that we're proud in past

  • Be part and take the graduation for Y-Combinator startup school 2020 (SUS2020) in March with this team;
  • Exceed the state of the art, percentage of accuracy for COPD and heartbeat classification;
  • Find the right compatible devices and start to building partnership with hardware supplies and Hospitals

What we learned

Probably managing Audio deep learning as a 1D time series, instead of transforming it in images (as done in the majority of the research reviewed) is effective, but requires a specialized network architecture.

That our team has the skills to turn the emergency around and is fun do hackatlon!

The solution’s impact to the crisis

Doctors will already have all the triage data in their database, this will help giving a quick glance to all their patients’ overall being and evaluate in no time whether to suggest them an in-hospital consultation.
Thanks to our efforts into looking for subjects at risk, we can prevent and propound the doctor who to mostly keep monitored.
We would have a much better healthcare system organization, a newer and quicker doctor-patient online relationship, and more focus on prevention to avoid the aggravation of clinical conditions that could have been treated immediately, beforehand.
A further positive impact falls on patients not affected by covid-19, but with other pathologies. With this type of organization, even non-covid patients will have the right care and attention from doctors.
The Mews score allows to evaluate vital parameters, discriminating stable, unstable or critical clinical conditions and is applicable to a wide spectrum of pathologies.
In conclusion, thanks to IntelliHearts even those suffering from other diseases can constantly monitor their parameters from home and share them with their doctor, allowing a more efficient management of all patients.

The necessities in order to continue the project

Imaging to the have high numbers of visitors on the tests of respiratory disease and Online triage we could need more and stronger servers.
So we need to get funded to pay servers, services, device’s certification and software medical certification, hire additional engineers and physicians. Support medical trials.

The value of our solution after the crisis

According to WHO estimates, 251 million people in the world have chronic obstructive pulmonary disease (COPD) and cardiovascular diseases (CVDs) are the number 1 cause of death globally, taking an estimated 17.9 million lives each year.
Thanks to IntelliHearts we will help patients with chronic obstructive pulmonary, cardiovascular, infectious and Metabolic diseases.
The project was born before the crisis for automatic diagnosis with a smartwatch and AI enabled app. Thus, our solution is ready for the post-crisis era, where people will pay more attention to their health and people understood the value of medical doctors, nurses and health operators, thus will reduce the accesses to family doctors (primary care physician) and hospitals by continue using new technologies (which they were forced to use during the lockdown) and managing everything remotely with remote health platforms.
We estimate that our technology can reduce by 40% hospital costs, playing an important role in prevention more than treat them and following the vision of current medicine.


Cardiac injury is a common condition among patients with COVID-19, and it is associated with a higher risk of in-hospital mortality.The findings presented in the research on 416 patients in Wuhan, Shi and colleagues [2], highlighted the need to consider cardiac complication in COVID-19 management. The patients with Covid-19 have respiratory distress and low blood oxygen levels, consequently they have high risk of ischemia or heart attack that compromises myocardial contractility and this situation can cause severe arrhythmia.
Respiratory diseases detection was conducted in the following way: We used the data from paper [1] Filter noise and remove background sounds We used several segmentation techniques based on spectrograms of respiratory audio Segmentation is necessary to understand when the first respiratory cycle starts. We used a custom deep learning neural network to predict binary healthy/unhealthy patients using only audio recorded by smartphone Developed a web service and web page to record and show the result. The accuracy is tested on the official test set in [1]
Why is early detection of COPD important for covid19?
SARS-CoV-2 uses the angiotensin converting enzyme II (ACE-2) as the cellular entry receptor to infect the lower respiratory tract. ACE-2 expression in lower airways is increased in patients with COPD and with current smoking [3] . ACE-2 is expressed in a variety of different tissues including both the upper and lower respiratory tract and myocardium. Importantly, nearly all deaths have occurred in those with significant underlying chronic diseases including COPD, and cardiovascular diseases These findings highlight the importance of increased surveillance of these risk subgroups for prevention and rapid diagnosis of this potentially deadly disease.


1- Rocha, B. M., Filos, D., Mendes, L., Vogiatzis, I., Perantoni, E., Kaimakamis, E., ... & Paiva, R. P. (2018). Α respiratory sound database for the development of automated classification. In Precision Medicine Powered by pHealth and Connected Health (pp. 33-37). Springer, Singapore.
2- Shi S, Qin M, Shen B, et al. Association of Cardiac Injury With Mortality in Hospitalized Patients With COVID-19 in Wuhan, China. JAMA Cardiol. Published online March 25, 2020. doi:10.1001/jamacardio.2020.0950
3- ACE-2 Expression in the Small Airway Epithelia of Smokers and COPD Patients: Implications for COVID-19. Janice M Leung, Chen Xi Yang, Anthony Tam, Tawimas Shaipanich, Tillie L Hackett, Gurpreet K Singhera, Delbert R Dorscheid, Don D Sin. Published in European Respiratory Journal doi: 10.1183/13993003.00688-2020
4 - Halpin, D. M., Faner, R., Sibila, O., Badia, J. R., & Agusti, A. (2020). Do chronic respiratory diseases or their treatment affect the risk of SARS-CoV-2 infection?. The Lancet Respiratory Medicine.

What's next for IntelliHearts

05/2020: Relase the platform letting people be able to use Machine learning detection and medical scores.

06/2020: Get the necessary medical certification.

07/2020: Involve hospitals for trials.

08/2020: find partnerships and investors.

Our Links:

Presentation link: IntelliHearts presentation for EUvsVirus 2020

Respiratory disease test:

Video of Cardiac Disease test Our AI Application that work with one of the compatible Smartwatch, running a cardiac analysis and parameters visualization

Video explain respiratory Disease test Our AI on respiratory audio

Covid Risk test (based on Italian Guidelines) Covid risk test online

GitHubGitHub Repo

Check out Our website to keep in touch and have more Info

Built With

Share this project:


Private user

Private user posted an update

For researchers reading this page, I have to add that accuracies are calculated using the medically proven inter-patient separation scheme for all the tasks: 1 - MIT BIH Arrhythmias database, the AAMI suggested which patients to use for training and which to use for testing. Accuracies that do not respect this separation are not counted towards the state of the art. 2 - AFIB the dataset is public and used in a conference paper competition, similarly to MIT-BIH conference owners suggested patients for training and other patients for testing. 3 - The respiratory disease audio database has a extremely unbalanced and difficult inter-patient separation scheme, original authors provided the patients to train on and the patients used for testing, the accuracy in state of art with the original separation scheme ranges fom 66% to 78% we hit 95% on the same train test split deeply respecting the inter-patient separation scheme.

Once again, in a race for accuracies , interpatient separation scheme is THE GOLD STANDARD when it cames to medical algorithms for CAD (computer aided diagnosis).


DISCUSSION: Algorithms with classical 70/30 split won't be kept into consideration and do not count towards the state of the art. Why? because using a mixed intrapatient separation schemes, beats or breath cycles of paople used for training will be used, partially, also for testing. This is called bleeding of information. Most importantly, the algorithm is not answering the question: is this person affected by this pathology ? but is answering another question: what is the owner of these beats and what is the pathology?

Log in or sign up for Devpost to join the conversation.