KIRETT - A wearable device to support rescue operations using artificial intelligence to improve first aid

Johannes Zenkert, Christian Weber, Mubaris Nadeem, Lisa Bender, Madjid Fathi, Abu Shad Ahammed, Aniebiet Micheal Ezekiel, Roman Obermaisser, Maximilian Bradford · 2025

Read KIRETT - A wearable device to support rescue operations using artificial intelligence to improve first aid by Johannes Zenkert, Christian Weber, Mubaris Nadeem, Lisa Bender, Madjid Fathi, Abu Shad Ahammed, Aniebiet Micheal Ezekiel, Roman Obermaisser, Maximilian Bradford (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-24934.txt.

Introduction

Johannes Zenkert and co-authors build a wrist aid for first care in a 2025 paper. At a crash or fire, first hands must act fast and right, and a wrong act can cost a life that right care would save.

Their aid is a worn tool with a smart core. It reads the scene, names what it sees, and cues the next act for the kind of harm at hand. The cues fit the case, not a stock list. The paper gives a first tour of the work: the aid's build, the lore it draws on, and the trials to come. The aim is plain: fewer false acts, more lives held, on the worst days a town can face.

Catalog record

CREATORJohannes Zenkert, Christian Weber, Mubaris Nadeem, Lisa Bender, Madjid Fathi, Abu Shad Ahammed, Aniebiet Micheal Ezekiel, Roman Obermaisser, Maximilian Bradford
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-24934

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

KIRETT - A wearable device to support rescue operations using artificial intelligence to improve first aid Thanks: 2022 IEEE International Smart Cities Conference (ISC2) | 978-1-6654-8561-6/22/$31.00 ©2022 IEEE | DOI: 10.1109/ISC255366.2022.9922361. Johannes Zenkert, Christian Weber, Mubaris Nadeem, Lisa Bender, Madjid Fathi, Abu Shad Ahammed, Aniebiet Micheal Ezekiel, Roman Obermaisser, Maximilian Bradford Abstract This short paper presents first steps in the scientific part of the KIRETT project, which aims to improve first aid during rescue operations using a wearable device. The wearable is used for computer-aided situation recognition by means of artificial intelligence. It provides contextual recommendations for actions and operations to rescue personnel and is intended to minimize damage to patients due to incorrect treatment, as well as increase the probability of survival. The paper describes a first overview of research approaches within the project. K eywords Rescue Operations \cdot Situation Recognition \cdot Artificial Neural Network \cdot Knowledge Graph \cdot Wearable Device 1 Introduction Rescue personnel worldwide are under extreme pressure and have to make decisions in the shortest possible time which can make a difference between life or death. For them, it is necessary to recognize relevant health situations of emergency patients on site and to take appropriate first aid measures. It happens often that an emergency patient is facing drastic changes in health vitals for an unrecognized reason, which results in certain modifications of emergency treatments. That is why it is necessary to find a fast and reliable solution for assessing the patient’s condition in real time and provide necessary treatment in rescue scenarios. The ’KIRETT’ project is developed to support the rescue personnel by means of a wearable that will be used for situation recognition, context-based treatment recommendations and minimizing potential damage due to incorrect treatments, to increase the probability of survival and lower the risk of long time impairments. To improve the quality of care and to increase the efficiency of rescue operations, an artificial neural network (ANN) algorithm will be developed to recognize the patient’s health situation. Afterwards, using a knowledge graph, the proper treatment path will be advised to the rescue personnel through the wearable. ANN is often used by current researchers to identify health complications like diabetes, cancer, asthma or cardiac arrest. AI based disease detection algorithms proved to be an effective tool for correctly diagnosing these health complications and creating numerous opportunities for reducing human errors, improving clinical outcomes, tracking data over time, etc. [ 1 ] . In the given research study, the developed ANN model will provide a vector output indicating the probability of complications that can exist when a rescue attempt is initiated. The model will be trained with records of the past five years from the rescue station from Siegen-Wittgenstein in Germany. Before the training, the raw dataset from the rescue station will be sorted and filtered using open-source data management system and Python programming. All the data modifications and analysis tools development including knowledge graph is done in local computer. The rescue station will also provide the literature of treatment instructions that consists of standard operating procedures in rescue services, as well as the guidelines for treatment and expert knowledge. Text mining will be used as an analysis method to extract information from the literature and other structured and unstructured knowledge sources. The extracted information will be stored in the knowledge base as dimensions, which can be used among other things, for the construction of knowledge graphs to display recognized connections or linked content or as a basis for the inference of new knowledge. 2 Situation recognition through neural network architecture A neural network will be used to recognize likely emergency health situations e.g., cardiac arrest, respiratory disease, pulmonary embolism of a patient. The ANN algorithm will be trained with more than 300,000 past rescue records consisting of health vital information, the primary diagnosis, first aid plans and rescuers’ impressions. In the further course of patient’s treatment, the situation may change afterward as the rescue forces assessing this primarily on the basis of the visible condition of the emergency patient, the data from the medical equipment (e.g. ventilator, ECG) and the mission description from the control center. For example, complications such as a reduction in respiratory drive during ventilation or cardiac arrhythmia during a heart attack can occur. Such situations must be responded to immediately, for example by resuscitation or appropriate medication. The developed neural network will help to make a quick decision and save the patients by continuously interpreting their situation based on health vitals and access to more data sources as soon as the equipment is attached in first aid situation. 2.1 Identification of the complications and their attributes The first step in developing the neural network is to create a training data set which consists of multiple output labels. The labels represent the complicated disease groups that can occur in an emergency situation. In collaboration with the rescue station, Siegen and professional medical partners, the possible complication groups to be labelled as ANN output are identified. From the rescue station database, it is found that the most common complications occur in emergency situations are pulmonary diseases, CNS diseases, cardiovascular diseases, respiratory diseases, abdominal diseases, psychiatric diseases, metabolic diseases, gynecologic-obstetrical emergencies, infections, and some other special diseases. For each complication group, such as cardiovascular and neurological disease, a thorough research was conducted to find out the most relevant attributes which are most common as disease symptoms and can be used to identify the complications. The attributes are mostly health vitals like blood oxygen saturation (SpO2), systolic and diastolic blood pressure, blood sugar level etc. Furthermore, text attributes containing important health information or diagnostic information from rescue personnel are categorized based on the severity, disease relevance and expert opinion. The complication groups are composed of multiple diseases. The current approach is to first identify the patient’s situation with a particular complication group, but not a particular disease that falls under that group. This is due to the number of total diseases inside all those groups being large in quantity and if used as output labels for the ANN may cause overfitting and high output latency. 2.2 Data sorting and filtering The database of past rescue attempts consists of thousands of unique cases identified with numerous complication types. Each of those cases was primarily recorded by the first responder at the rescue station with the relevant important information: Parameters of the rescue, geographical information of the operation, patient’s medical history, health diagnosis, vital signs, medications administered and treatment path performed. The database handed over for research was in raw form. Due to the manual documentation by the experts, not every case was provided with sufficient or complete information. Furthermore, much of the information recorded in the database was irrelevant for later analysis. So, a filtering algorithm is created to limit information, remove outliers and extrapolate some data fields when further knowledge sources are available. A data dictionary software is created that makes a comprehensive analysis of the nature of the selected attributes and supports an on-demand data composition and export. It furthermore stores descriptions of all acronyms and data types to support the collaboration between data engineers. The software is written using Python version 3.9 and the

Read the complete text