AI: emergency services’ best sidekick?
- March 17, 2021
- Categories: Artificial intelligence
We will start by understanding the basics of this matter. What is the difference between dogs and bagels? I know what you’re thinking…What do dogs have to do with this famous bread? And what does all of this have to do with AI? The example below will help you to understand.
The images above show us that, when distinguishing the differences between images (in this case bagels and sleeping dogs), the error rate of humans remains the same over time. However, algorithms learn and significantly decrease their error rate.
If in the beginning algorithms confused similar images, now machines outdo a human’s performance when recognising objects, with an error rate lower than the 5% human rate.
This principle can be utilised in other fields like text analysis or speech recognition. Therefore, using algorithms to automate certain tasks and reduce our risk of error can bring considerable help to emergency services.
Before we continue, it is important to understand that this article is not about replacing humans with intelligent machines, but about the creation of a hybrid space where humans and machines can work together to improve the emergency response sector.
When dealing with an emergency, there are numerous external factors that can affect a decision. Physiological and environmental pressure or even emotions can hinder our rationality and affect our judgment in a life-or-death situation.
Machines can process huge amounts of data faster than humans can. They can also analyse a situation detached from emotions and as seen above, learn to constantly improve.
Public safety organisations must be open to this new technology which must be accompanied by a robust ethical framework and be in compliance with privacy regulations.
Want to know more about AI and emergency services? Watch our sessions from EENA Conference 2018 (here) and from EENA Conference 2019 (here).
AI/ML and emergency calls
Besides processing data, AI and ML can also find patterns and new insights.
Emergency call centres are often swamped with emergency calls. These tools could help analyse calls and texts, which would ultimately reduce waiting times and save more lives.
According to a recent study, emergency call-takers fail to identify around 25% of cases of out-of-hospital cardiac arrest and consequently lose the opportunity to provide CPR instructions to the caller.
Every year, over 350,000 cardiac arrest cases occur outside of the hospital setting. If CPR is applied within the first minutes, the survival rates increase substantially.
AI could fill this gap and help with the early detection of cardiac arrest, and this is what the Danish company Corti is working on.
This company developed an AI system that analyses emergency calls to predict out-of-hospital cardiac arrests more accurately. The system can detect a cardiac arrest based on historical data. The algorithm listens to the call in real time and will alert the call-taker on the screen if the caller is having a cardiac arrest.
In 2018, EENA launched a pilot project with Corti to test how this system could work alongside emergency call-takers in different pilot sites in France and Italy. Machine learning proved to be an important tool to support these professionals. The study concluded that a machine learning framework could augment the capacity of call-takers and dispatchers to identify out-of-hospital cardiac arrest in emergency calls.
AI/ML and injury diagnosis
AI has been successfully used in diagnosing injuries.
The Israeli start-up MDGo is using advanced AI technology to help dispatchers to know if a car accident requires an ambulance.
When a car crash occurs, their system creates a medical report in real time with data regarding the forces applied on the passenger (e.g., duration, moment, vector). This data is generated by their algorithm and sent automatically to the Israeli emergency medical services, Magen David Adom (MDA). This tool can bring first responders to the scene quicker and prevent complications from injuries that otherwise might not be identified.
When a person is in distress, according to this article, “it takes an average of five minutes during the day and seven minutes at night before someone calls for an ambulance”. With the MDGo system, the emergency services are alerted in seconds and ambulances can be dispatched automatically according to the type and severity of the occupant’s injuries. With this AI-based technology, the start-up estimates that auto fatalities can be reduced by 44%.
AI/ML and firefighters’ safety
AI can also help firefighters to respond more effectively and reduce exposure to extreme conditions.
The Spanish start-up Prometeo developed an AI-based cognitive health monitoring platform to measure toxins that firefighters are exposed to when battling fires.
Bee2FireDetection, a solution developed in Portugal, uses Visual Recognition AI on captured images from surveillance cameras and other data sources to detect fires in forests, natural parks and even mining areas and industrial facilities. This early fire detection is fundamental in containing or reducing damage, monitoring potential risk areas, and shortening reaction times.
This solution was implemented in Europe, the United States and Brazil (Amazon region).
AI/ML and crisis management
In disaster response, it is crucial to have a good grasp on what a new terrain looks like. AI can help disaster responders to know what resources need to be mobilised. The violent Nepal earthquake in 2015 demonstrated the practical use of AI to spot urgent needs and identify infrastructure damage.
Another example is the Center for Robot-Assisted Search and Rescue. This organisation has been providing support during several disasters, from man-made incidents to natural hazards, by fostering unmanned systems that are being used effectively by formal management agencies.
Their work and the work of other organisations has proved that AI can support relief efforts and change the way emergency teams respond to and prevent a disaster. As an example, they deployed humanitarian aid with a robot lifeguard to a Syrian boat of refugees arriving to Greece to understand which people needed immediate help. The organisation also explained the importance of AI to help identify which language a person uses and the good use of ML to enhance situational awareness (e.g. predict the wind, assess how many boats will arrive on the shore, etc).
Another interesting initiative was conducted by the Humanitarian OpenStreetMap Team (HOT). As part of Microsoft’s AI for Humanitarian Action programme and in partnership with Bing Maps, the team produced AI-detected open building datasets for Tanzania and Uganda to meet the challenge of mapping unmapped areas and improve mapper experience. Mapping is fundamental in disaster risk reduction and integrating deep learning can be very valuable when developing a sturdy humanitarian response.
In an emergency, situations change very quickly. Above, we highlight just a few examples of how AI can be a game changer for public safety. This includes helping to get information faster and more accurately, reduce repetitive and procedural activities currently carried out by humans, identify new trends, or even increase visibility and understanding.
Implementing AI in emergency response requires great consideration. AI holds a vast potential if implemented correctly – considerations about ethics, fairness and compliance with privacy regulations should be prioritised.
AI is not a panacea but can for sure be a good sidekick to help emergency services professionals by augmenting and enhancing their work. It is time to stop being afraid of a human-machine collaboration. The technology is already being used in other sectors of society and is getting closer and closer to having a predominant role in the public safety sector.
If emergency services are not using these technologies, which are already used by other sectors, this could lead to opportunities being missed and, consequently, lives being lost.
If UberEATS uses ML to optimize deliveries, wouldn’t it be worth emergency services using it to find you in a crisis?
If Facebook uses AI to personalise your newsfeed and ensure you see only what you are interested in, wouldn’t it make more sense to also use this technology to help call-takers respond quickly?
AI/ML is everywhere, from the spam filter of our email inbox to the selection of the preferences of our social media feeds. Therefore, it would be a shame to continue to resist the use of this technology in a way that can save lives.
References:
American Heart Association (n.d) CPR Facts & Stats – How CPR is changing (and saving) lives. American Heart Association. Retrieved from https://cpr.heart.org/en/resources/cpr-facts-and-stats
Blomberg, Stig & Folke, Fredrik & Ersbøll, Annette & Christensen, Helle & Torp-Pedersen, Christian & Sayre, Michael & Counts, Catherine & Lippert, Freddy. (2019). Machine learning as a supportive tool to recognize cardiac arrest in emergency calls. Resuscitation. 138. 10.1016/j.resuscitation.2019.01.015.
Cleve et all (2020) Project Report: Detecting out-of-hospital cardiac arrest using artificial intelligence. EENA. Retrieved from https://eena.org/knowledge-hub/documents/detecting-out-of-hospital-cardiac-arrest-using-artificial-intelligence/
Gomez et all (2019) Artificial Intelligence & Machine Learning in Public Safety. EENA. Retrieved from https://eena.org/knowledge-hub/documents/artificial-intelligence-machine-learning-in-public-safety/
Leichman, A. (2019) Now your car can call an ambulance for you. ISRAEL21c. Retrieved from https://www.israel21c.org/the-man-who-meets-indias-business-needs-with-israeli-products/ Patrick Meier (2015) Virtual Aid to Nepal. Foreign Affairs. Retrieved from https://www.foreignaffairs.com/articles/nepal/2015-06-01/virtual-aid-nepal
Author: Marta Azevedo Silva, Former Communications and Press Manager at EENA
The opinions expressed are those of the author and do not necessarily represent the views of EENA. Articles do not represent an endorsement by EENA of any organisation.
EENA
Share this blog post on: