Triage.io

• 

  Interface for Triage.IO

Problem

The COVID-19 outbreak affects all segments of the population and is particularly detrimental to members of vulnerable communities. The pandemic continues to affect populations, including people living in poverty, elderly people or those with disabilities, youth, and indigenous people. The current pandemic has an unprecedented impact on Medical AID services and Government budgets, while the social distancing, self-isolation and economy impact are expected to induce radical changes in the social behavior of citizens everywhere. Medical AID services have been overwhelmed with triage and treatment, being unable to ensure secure live triage for contagious cases. The crises impacted mostly elderly people, those unable to access immediate medical care when the first signs and symptoms were noticed. While medical services have been hugely impacted by the infectious cases, we have witnessed enormous effort of central and local governmental entities to monitor and triage all quarantined citizens. The result is huge cost on the government level in each country. On the economy level, as a future prospect, companies will need to take unprecedented measures to further protect their workforce from employees with suspicious symptoms.

Solution

Triage.IO with use latest technologies to ensure instant 24/7 citizens medical evaluation, without further involvement of the medical-triage dedicated personnel and operational costs. The solution uses Symptoms Extraction & Evaluation AI Engine from voice, rich media & IVR ( Interactive Voice Response) , chat, text-files or X-RAY scanning when available. The solution can perform a Virtual Triage Room for the entire medical system, that will offload medical services, optimizing Human Resources involved in the symptomatology evaluation. All medical personnel will no longer be exposed to suspicious cases without prior notification. The system will allow online daily evaluation for quarantined and isolated cases, identification of scattered cases, patient tracking and will provide online historically records to perform population analytics, live forecasting of potential diseases and instant outbreak alerts.

Technical Solution

Triage.IO uses models for the AI Engine components, fine-tuned for medical vocabulary:

Transformer Neural Network

• Automatic Speech Recognition
• Human-Level Text-to-Speech with proper accent, intonation, customizable timber of voice.
• Bidirectional Encoder Representations for Natural Language

Understanding based of BERT language representation: o Relevant entity extraction from text and speech o Intent Classification o Sentiment Analysis

Xtreme Gradient Boosting technology for feature-aggregate classification.

State of the art Convolutional Neural Networks for imagistic analysis

Why Triage.io

We identified the following scenarios that the solution will bring benefits by increasing the ability to fast evaluate and triage suspicious cases :

COVID -19

• Protection of healthy citizens and medical personnel in triage rooms
• Removing the need for medical human resources in first line of contact, diminishing contagion risk
• Helping people from isolated communities, people with disabilities & elderly people reach evaluation services

POST COVID -19 / RECOVERY STAGE

• Interactive Voice Response (IVR) support for patient-care centers in hospitals
• Forecasting recurrent outbreak of infection
• Pre-screening / medical evaluation for telemedicine applications
• Transcript support for telemedicine application
• Voice Assisted imagistic reports
• Fast identification of mutated symptomatology

WORKFORCE HEALTH CARE

• Corporate daily enforcement of workforce health-status checkin

DEMO Triage.io

During the Hackathon we will implement a Proof of Concept App by using a state of the art BERT English pre-trained language model, fine-tuned for extraction of relevant natural language voice/text symptomatology-input.

The following scenario will be implemented:

User – Citizen CONNECTS to the application

User DECLARES symptoms using VOICE

Triage.IO EVALUATES symptoms and PROPOSES Triage-Flag

What's next for Triage.io

The current technology is already fine-tuned for medical vocabulary in English and Romanian languages. A large scale implementation with the development of the application, integration with medical emergency services will take 45-60 days, including the fine-tuning for EU languages.

Built With

• html
• javascript
• python
• tensorflow