Presentation of the Kid AID Project - AI-Powered System for Non-Invasive Assessment of Child’s General Condition Supporting Early Sepsis Detection
Speaker: Kamila Ludwikowska (Poland), Monika A. Tokarczyk (Poland), Adam Łęcki (Poland)
In this webinar, international speakers explored AI-enabled solutions for improving early recognition and triage of pediatric sepsis. The discussion centred on the gap between clinical sepsis definitions and real-world diagnostic tools, proposing innovative, data-driven solutions to enhance frontline decision-making.
Sepsis Definitions and Diagnostic Gaps:
Speakers outlined the limitations of current pediatric sepsis definitions (e.g., Phoenix Sepsis Criteria), which are often retrospective and incompatible with frontline workflows. The inadequacy of available bedside tools and the subjective nature of clinical assessments contribute to diagnostic delays and variability. An illustrative anecdote recounted a seasoned nurse's intuitive identification of sepsis—highlighting the value of tacit clinical knowledge that AI may help replicate and standardize.
Clinical Case Illustration:
The webinar featured a real-world case involving a child who presented without fever and had nonspecific symptoms. A nurse, relying on clinical intuition, suspected sepsis and initiated early intervention. The child was later confirmed to have Neisseria meningitidis and deteriorated rapidly. This case underscored the limitations of standard protocols and the value of experience-driven assessment, reinforcing the need for tools like KidAID to support early recognition.
KidAID Platform:
To address these gaps, the KidAID project was introduced—an international research initiative leveraging real-world data and video-based assessments to build scalable AI tools. A central component is KidAID, an AI-augmented triage system combining clinical videos, sensor data, and machine learning. It uses a standardized 1-minute video protocol, where a caregiver responds to guided prompts while the child is recorded using a non-invasive, dual-screen device. One screen displays child-friendly content to maintain attention; the other is a clinician interface.
The tool classifies risk using a traffic light triage model (green/yellow/red), enabling early warning flags before clinical deterioration. Data is processed securely, with full encryption and privacy safeguards.
Augmented Intelligence, Not Replacement:
Speakers emphasized AI's role as “augmented intelligence” and not clinician replacement. One speaker described it as a "stethoscope with a neural network," underlining the system’s potential as an extension of clinical expertise rather than a substitute. The tool is designed to assist, not override, human judgment—particularly in ambiguous cases.
AI Training and Validation:
The training process involves data cleaning, manual annotations, and feature extraction, followed by model training using a 75/25 data split for validation. All data is processed through secure cloud infrastructure, supporting scalability and real-time analysis.
KidAID Coach – Education and Training:
The KidAID Coach module aims to train healthcare workers using actual patient videos. It supports multilingual and cross-cultural settings, with standardized response formats to accelerate knowledge transfer across geographies. This educational component ensures that the tool can be adopted equitably in diverse health systems.
Early Adoption and Real-World Testing:
KidAID has been deployed in hospitals in Poland and is being piloted in Saudi Arabia. Early results show promise in improving triage accuracy and reducing time to intervention.
Audience Engagement:
Live polling showed strong support from participants for integrating AI into pediatric emergency care. The audience endorsed AI as a tool that could enhance—but not replace—clinical decision-making, particularly in high-risk, time-sensitive scenarios.
Conclusion:
The webinar showcased an evidence-informed vision for AI in pediatric emergency medicine. Through video-based triage, cross-border collaboration, and AI-enhanced education, the KidAID platform aims to close the gap between sepsis recognition and clinical response. Its design reflects a thoughtful balance of technological innovation, clinical realism, and ethical responsibility, offering a promising path forward in the global fight against pediatric sepsis.
ESPID 2025, 26-30 May, Bucharest
