Mawasqat: AI for Symptoms, Diagnoses, Medications and Lab Data

AI for Symptoms, Diagnoses, Medications and Lab Data

This Mawasqat product uses clinical NLP and speech-to-text to turn unstructured clinical speech and text into structured, interoperable data for EHRs, analytics and downstream AI systems.

Healthcare AI
AI & machine learning
Product

The Challenge

What stood in the way.

Clinical information often starts as speech or free text. Turning it into structured, interoperable data is difficult, because symptoms, diagnoses, medications and lab tests are described in many different ways.

Unstructured text and speech

Notes and conversations do not arrive in a form EHRs and analytics tools can read.

Varied speech

Speakers and accents differ, which makes accurate transcription harder.

Interoperability

Structured output has to follow healthcare standards so other systems can use it.

Our Solution

How we answered it.

We built an AI pipeline that combines multi-model speech-to-text with clinical NLP. It detects symptoms, extracts ICD-11 diagnoses and medical entities, understands medications, identifies lab tests and summarises encounters

ProblemClinical speech varies between speakers.
Our answerWe combined multiple speech-to-text models with accent normalisation.
ProblemFree text hides the clinical facts.
Our answerWe built extraction for symptoms, ICD-11 diagnoses, medications, lab tests and other medical entities.
ProblemStructured output still has to be usable elsewhere.
Our answerWe mapped results to FHIR and produce EHR-ready outputs.

How It Works

The parts that made the difference.

Symptom detection from EHR text

The model detects symptoms mentioned in EHR text.

Multi-model speech-to-text

Several speech-to-text models are used to transcribe clinical speech.

ICD-11 diagnosis extraction

Diagnoses are extracted and coded to ICD-11.

Medical entity extraction

Medical entities are pulled out of the text as structured fields.

Results

What changed for the client.

The product turns unstructured clinical input into data other systems can use.

Unstructured clinical speech and text converted into structured, interoperable healthcare information
Output ready for EHRs, analytics and downstream AI systems
Diagnoses coded to ICD-11 and data mapped to FHIR