Method for pre-diagnosis in emergency situations using artificial intelligence

Abstract

The present invention relates to a method for pre-diagnosis in emergency situations, which is based on computer vision-type artificial intelligence. By analysing the images sent by caller smartphones, a pre-diagnosis is supplied to the operators of emergency call centres, covering three major aspects: traumatology, nature of the emergency and emotional status of the patients. The method makes it possible to assess the nature and criticality of the patient's injuries, to determine the most suitable emergency department according to the context of the emergency, and to estimate the level of fear and stress of the patients. The innovative method automatically makes it possible to qualify, analyse, sort and pre-diagnose emergency calls on the basis of images sent during emergency situations. Beyond the use to improve the diagnosis of everyday emergencies, the solution becomes a vital tool against large-scale crises that generate a number of emergency calls that is impossible to process with the human resources of call centres.

Description

[0001] Description

[0002] Title of the invention: Method for pre-diagnosis during emergency situations using artificial intelligence

[0003] The present invention relates to a method for pre-diagnosis during emergency situations based on artificial intelligence of the "Computer Vision" type. Analyzing the images transmitted by the callers' smartphones (using an application or a cloud service), a pre-diagnosis is provided to the operators of emergency call centers (SAMU / Firefighter / Police), following three major axes: trauma, nature of the emergency situation and patient emotion. ("AI HighWind")

[0004] In France, approximately 69 million emergency calls per year are made to the SAMU (Emergency Medical Services), the Fire Brigade, and the Police. They are based on GSM technology (voice phone calls), which has not undergone any major developments since the introduction of 112 in 1997 with the first mobile phones. The few initiatives to use smartphone capabilities operate at the local level only, mainly for the inclusion of GPS position or videoconference links transmitted by SMS. This lack of technology in emergency calls results in numerous errors and underestimations of the severity of the situation, leading to many deaths each year. Indeed, the emergency operator must assess the criticality of a situation based on a verbal description from the caller, lasting between 10 and 15 minutes out of the 60 minutes of the "Golden Hour" that allows a life to be saved during a serious emergency.Furthermore, faced with large-scale crisis situations (natural or industrial disasters, terrorist attacks, fires, etc.), emergency call services are overwhelmed by the exponential number of calls (>100 calls per minute compared to approximately 2 per minute in normal times) and are no longer able to distinguish the criticality and nature of the calls. These crisis calls are generated not only by patients closest to the epicenter of the crisis but also by surrounding people who are not facing imminent danger, the current method of sorting emergency calls not allowing them to be distinguished without spending significant time talking to each caller individually. The method according to the present invention makes it possible to overcome this drawback.Indeed, HighWind, a French start-up, has developed an emergency call solution that allows photos, videos, GPS position, pre-filled information and VoIP of an emergency situation to be transmitted via the smartphones of the target population, in addition to the traditional call, to emergency call centers equipped with a HighWind interface.

[0005] The solution intervenes at several levels:

[0006] - Public: within the caller's mobile phone, through an emergency call smartphone application or a cloud interface opened by SMS link.

[0007] - AI HighWind: data analysis to provide a pre-diagnosis of the emergency call using the HighWind process using artificial intelligence of the type

[0008] “Computer Vision”.

[0009] - Emergency call centers: via a dedicated interface (software or web) allowing emergency call operators to view the information sent by callers, enriched by HighWind's AI analysis.

[0010] HighWind therefore provides an emergency call service (“HighWind Service”) through an emergency call solution (“HighWind Solution”) which integrates all the elements of transmission, analysis and visualization of emergency information, from the caller via their smartphone (“HighWind Application”) to the emergency call center which receives it (“HighWind HQ Interface”) pre-analyzed by artificial intelligence using the innovative technological process developed (“HighWind AI”).

[0011] Whether for a public or private population (e.g., in a company), the HighWind emergency call application or cloud service is always distributed free of charge. The HighWind Application also allows, thanks to its internal artificial intelligence, to call emergencies anywhere in the world, whether the emergency call center is equipped with the HighWind HQ interface or not.

[0012] The HighWind Solution exploits the tunnel effect that callers experience during an emergency situation, the restriction of their visual field pushing them to create a Situation-Phone-Eye alignment which allows photos to be taken (main and rear "selfie") by a single click on a single button of the HighWind Emergency Call Application.

[0013] The objective of the HighWind Solution is to increase the chances of survival of patients, in public life or in business, by significantly improving the precision, accuracy and speed of emergency diagnoses. Faced with emergency calls lasting on average 10 to 15 minutes within the 60 minutes of the "Golden Hour" of serious emergencies (the risk of death increases exponentially with the passage of time), the HighWind Solution is capable, by reducing the duration of an emergency call, of improving the chances of survival of patients. The present invention relates to a pre-diagnosis method during emergency situations, based on the recovery and rapid processing by artificial intelligence of the "Computer Vision" type of images of emergency situations provided by the callers' smartphone or other means of capture. Said method delivers an analysis along three major axes: a. The nature and criticality of the trauma; b.The nature of the emergency situation associated with determining the most appropriate emergency service; c. Estimation of the patient's emotions.

[0014] The pre-diagnosis process makes it possible to provide an emergency reception center (public: emergency call centers such as SAMU / Police / Firefighters, but also private: in companies or as a remote solution during major events), with images of pre-diagnosed emergency situations and to send analyzed elements allowing for determination, sorting, redirection, and assistance in the decision of the emergency operator in order to improve the patient's care and their chances of survival.

[0015] As a result, the pre-diagnostic process makes it possible to significantly reduce the time required to understand an emergency situation, either automatically or for the benefit of emergency operators.

[0016] The accompanying drawings illustrate the present invention:

[0017] Figure 1 is a visual of the concept of the HighWind Application distributed to the population.

[0018] Figure 2 is a visual of the type of architecture deployed at the heart of the process. The architecture used to deliver the HighWind Service is multiple and scalable, but follows the guiding principle of emergency pre-diagnostic analysis using "Computer Vision" type artificial intelligence. As an illustration, the model currently uses a dual architecture combining different means based on "Computer Vision" and used to support the HighWind Solution.

[0019] Figure 3 is a graph illustrating the performance of the method, and more specifically the fact that the present invention secures a pre-diagnosis on the nature and criticality of a traumatic injury at 85% in less than 90ms on open injuries. For comparison, it takes: 50ms for the eye to focus on an image so that it is transmitted to the human brain; 300ms for brain activity to increase considerably in order to analyze an image; several seconds or even minutes to determine the nature of a medical image; 10 to 15 minutes currently to provide a diagnosis of emergency calls with the telephone resources of the SAMU.

[0020] Figure 4 illustrates the pre-diagnostic chain enabled by the present invention (first step). Figure 5 illustrates the decision chain enabled by the present invention (second step).

[0021] In its current form, the HighWind App's step-by-step operation is as follows. It was established through feedback from emergency personnel, patients, and survivors of major crises (Figures 4 and 5):

[0022] 1. [Human] Open the app

[0023] Hidden [App] Confirm region / country

[0024] Hidden [Brain Reflex] eye-phone-situation alignment

[0025] 2. [Human] Click the call button

[0026] Hidden [App] Evaluates network speed (2G to 5G)

[0027] Hidden [App] Checks whether the SAMU center is equipped with HighWind or not

[0028] Hidden [App] Sends GPS location and pre-filled information

[0029] Hidden [App] Takes a main photo

[0030] Hidden [App] Takes a front photo (selfie)

[0031] Hidden [App] Sends photos to HighWind and SAMU

[0032] Hidden [App] Call 15 / 112 or the equivalent depending on the country

[0033] Hidden [App] SAMU receives pre-diagnosed images

[0034] 3. [Human] Talk to the emergency services on the phone

[0035] The behavior of callers was tested and with the feedback from rescuers, due to the tunnel vision effect and the need for situational awareness, the brain's mechanism always produces an alignment: Eye - Phone - Situation when dialing or using the HighWind Application. (Figure 1)

[0036] The unique and innovative nature of the HighWind Service lies in its emergency pre-diagnosis system.

[0037] HighWind AI defines the process of retrieving images of emergency situations transmitted by the caller's smartphone, analyzing them using "Computer Vision" type artificial intelligence and transmitting the analysis results and pre-diagnostics to emergency call centers (public, private, or any means of retrieving emergency calls). (Figures 4 and 5)

[0038] The process allows analysis according to 3 defined pillars:

[0039] - Traumatology: it allows the identification of the type of injury and its level of criticality, allowing in particular the recognition of hemorrhage, lacerations, bullet impacts, burns (fire, acid, electricity), open fractures, dislocations, ulcers, hypothermia etc.

[0040] - Emergency situations: it allows to recognize the nature of the emergency situation and determine the most appropriate emergency service, distinguishing for example: road accident, physical injuries, fire, landslide, earthquake, flood etc.

[0041] - Context & emotions: it allows us to determine the patients' feelings, particularly in terms of pain, fear, stress, muscle tension, body lying down, etc.

[0042] The results of the data analyzed along the three major axes thus make it possible, by artificial intelligence, according to the centers of interest of the emergency services, to provide different types of recommendations, analyses and emergency decision-making aids, which may include, for example, but are not limited to: the criticality of the patient's condition, the most suitable emergency service, the risks incurred by patients and rescuers (ballistic, blunt, natural, fire, slipping, etc.), the nature of the danger present, the priority between the different situations reported by callers, recommendations regarding the most suitable equipment for the intervention.

[0043] The three pillars pave the way for a fourth pillar enabling the automation of emergency call transcription and their future automated translation. Visually understanding the situation makes it easier to determine the vocabulary that will be used in a given language and significantly increases the technology's effectiveness.

[0044] As part of the issue of grouping emergency call numbers around a single number, brought to the National Assembly in a bill, on 2 ème pillar of HighWind AI technology, determining the nature of the emergency situation, allows by extension to determine which emergency service is most suited to a situation (SAMU, firefighter, police, etc.).

[0045] The artificial intelligence at the heart of this invention is of the “Computer Vision” type, using “Deep-Learning” training on medical databases, images of emergency situations and the use of the HighWind Solution, taking into account GDPR considerations.

[0046] The founding team of IA HighWind has been working since mid-2019, following discussions with SAMU 75, SAMU 06 and the Fire Brigade SDIS 13, on the development of an emergency call solution on smartphones allowing the communication of photos, GPS position, video and VoIP to SAMU centers equipped with the HighWind Solution. The present invention is therefore intended to serve emergency service operators (SAMU, SDIS approved civil security associations or companies specializing in telemedicine and emergency assistance for businesses), via the HighWind Solution.

Claims

AMENDED CLAIMS received by the International Bureau on May 15, 2024 (15.05.2024) 1. Method for assisting with triage and pre-diagnosis during emergency situations for SAMU, fire brigade and police call reception and dispatch centres, based on the recovery and processing by artificial intelligence of the “Computer Vision” type of images of emergency situations provided by the callers’ smartphone or other means of capture, followed by the analysis of the data collected by an artificial intelligence algorithm, to determine the nature and criticality of the emergency situation. Said method is characterised by the following steps: Taking images via the web application or downloaded; Rapid recovery and processing of emergency situation images by artificial intelligence of the “Computer Vision” type, using “Deep-Learning” training on medical databases, images of emergency situations, and seeking to recognize elements along 3 key axes: o TRAUMATOLOGY: it allows to identify the type of injury and its level of criticality, making it possible in particular to recognize hemorrhage, lacerations, bullet impacts, burns from fire, acid or electricity, open fractures, dislocations, ulcers, hypothermia etc. o SITUATION AND / OR CONTEXT: it allows to recognize the nature of the emergency situation, for example: road accident, physical injuries, fire, landslide, earthquake, flood etc.o EMOTIONS: it allows quantification of the level of pain expressed by the patient via “Deep-Learning” training on “selfie” faces from medical databases, images of emergency situations. Analysis of data by artificial intelligence, data from “Computer Vision”, making it possible to provide a pre-diagnosis of the nature and criticality of the emergency call. Visualization of information sent by callers, enriched with analysis by artificial intelligence of the “Computer Vision” type, within the emergency call center allowing triage, redirection, assistance in the decision of the emergency operator, display of the risks incurred by patients and AMENDED SHEET (ARTICLE 19) rescuers, and the priority between the different situations reported by callers, in order to improve patient care and their chances of survival.

2. Method for assisting with triage and pre-diagnosis during emergency situations for SAMU, fire brigade and police call reception and dispatch centres, based on the recovery and processing by artificial intelligence of the “Computer Vision” type of images of emergency situations provided by the callers’ smartphone or other means of capture, followed by the analysis of the data collected by an artificial intelligence algorithm, to determine the nature and criticality of the emergency situation, according to claim 1, characterised by: Determining the most suitable emergency service according to their activity, SAMU, firefighters, police, by recovering and processing by artificial intelligence of the “Computer Vision” type of images of emergency situations provided by the callers’ smartphone or other means of capture, via the 2 edetection axis of claim 1 “situation and / or context”, which allows recognition by type training “Deep-Learning” the nature of the emergency situation such as a road accident, physical injuries, fire, landslide, earthquake, flood. AMENDED SHEET (ARTICLE 19)

Images