endorsedCompleted

AI Medical Decision Support System for Thromboelastography

Services Covered:

Custom Software Development • AI and Machine Learning

Industry

Engineering & Manufacturing

Duration

2 months

Budget

10K - 49K

Start Date

1 October 2018

Team Size

1-5 Employees

Engagement Model

Offshore

Project Capability Score (PCS)

71/100

Good

Project Capability Score (PCS) estimates how capable Rubius is of successfully delivering a project like this, based on its past experience, track record, and ability to handle similar work. Learn more about PCS

Client

Mednord-Technic

Tomsk, Russia

Project Summary

Mednord-Technic engaged Rubius to develop an AI platform that automates interpretation of thromboelastography data so clinicians could obtain faster, information-based conclusions. Rubius processed 1,300 real test records, built a machine-learning classifier and delivered a Python-based medical decision support system that produces results in seconds and achieves 98.4% accuracy. The client praised the team's domain expertise, professionalism, and the project's rapid two-month delivery.

Key Challenges

Deciphering real-time thromboelastography graphs required time and highly qualified doctors
The client needed to simplify the interpretation process and make the methodology more accessible to specialists
The goal was to develop an AI system that could produce fast conclusions on a patient's hemostatic potential

Project Deliverables

Machine learning classification algorithm for blood clotting disorders
Medical decision support system
Data processing of 1,300 thromboelastography test records
Python-based implementation using NumPy, Pandas, Scikit-learn, Keras and TensorFlow

Project Solution

Rubius processed 1,300 real thromboelastography test records (each with 20 clotting indicators and a diagnosis), developed a machine-learning algorithm to classify blood clotting disorders, and implemented a Python-based medical decision support system. The implementation used NumPy, Pandas, Scikit-learn, Keras and TensorFlow. The vendor delivered the system through weekly sprints and a team that included a dedicated project manager and two developers, and prepared the solution for integration with the Mednord analyzer software.

Project Outcome

The system produces results in a few seconds
Achieved 98.4% diagnostic accuracy
Integrated into the Mednord thromboelastography analyzer software for bedside use
Enabled faster, more reliable diagnoses leading to correct treatment decisions

Platforms

AI/ML Platform

Tech Stack

Python
TensorFlow
NumPy
Pandas
Scikit-Learn
Keras

Client Endorsement

Overall Review Rating

Timeliness

Cost Rating

Willing to Refer

Quality of Deliverables

“The system generated by Rubius resulted in an outstanding accuracy rate of 98.4%. Thanks to this platform, doctors obtained a support system that allowed them to analyze information and make faster diagnoses. The client lauded the team for their field expertise and professionalism.”
Egor Zhukov
CCO
Date: 02/03/2021