logo

Personified Blood Test Analysis Solution for Remote Medicine

Published: November 11, 2022
# Healthcare
# Data Science
# NLP
We designed an ML-powered system for personified lab test results. It boosts the efficiency of physician office laboratories (POLs) and increases the speed of the medical service delivery system. Moreover, the system that we created became a part of a broader telemedical patient care delivery system. Read on to find out how we deal with challenges like this one.

Challenge

Traditionally, test results are compared against normal ones. However, “normal” parameters differ for each person due to age, sex, race, pregnancy, lifestyle, and related diseases. Moreover, it takes significant effort to digitize test results manually and keep all the data from disparate sources in the same place. Also, patients with chronic diseases and patients with a specific therapy take tests regularly to monitor their health state. Test results that every patient can understand helps to avoid confusion. In essence, we decided to deliver personification, digitalization, and clarity to our client.

Solution

  • First, we ensured the complete blood count (CBC) analysis. The incoming laboratory test results were uploaded and analyzed with the help of OCR and NLP techniques. Simultaneously, we developed a system for storage and long-term analysis (machine IQ).
  • Second, we created a connectionist temporal classification model (CTC ) with custom architecture based on the standard OCR example from Keras. Later we added the second module for blood metabolic panels. This stage included the development of a microservice-based API for laboratory integration and white label and an API for IoT and Health app integration.
  • Third, we added an ML solution for prediction and benchmarking to analyze the patient data and adjust the standard parameters to their medical history. So this is how understanding the “norm” for a specific person enables a more accurate prediction of future test results. Plus, it provides meaningful recommendations. Of course, we also developed an API for lab analysis machine integration.
  • Finally, we researched available telemedicine services to integrate the system with them. Integration with this service can provide high-quality, structured healthcare big data accessible for further research, marketing, and development.

Development Journey

Backend

Python, aiohttp, PostgreSQL, Redis, Docker, Kubernetes, Keras, OpenCV, spaCy

Frontend

AngularJS 6 Environment We are entirely relying on AWS: S3 for file storage CloudFront as CDN EKS as a managed Kubernetes environment for microservices EC2 for load balancing

Impact

Personalization: We created a system that provides individualized blood analysis results, considering every patient's medical history. As a result, the end-user gets precise results and can see any deviations. Digitalization: Test results are processed directly from a picture/scan or an email sent by the laboratories. All historical data is stored and used to track health to notify users of any changes. Clarity: All information is clear to the end-user and goes with recommendations regarding any further medical specialist's consultation or other tests.

So this is how you instantly provide personified delivery of patient care, crystal clear to the patients by having the lab tests and the patient's biomarkers.

RELATED CASE STUDIES

View all Case Studies