Regression Neural Network–Powered Platform for Real-Time Liver Graft Analysis

About Our Client

The Client is a medical device manufacturer headquartered in Europe. For over 20 years, the company has been designing and delivering medical technology for organ preservation.

Need for Secure, Data-Driven Platform to Support Liver Transplant Decisions

The Client launched an R&D initiative to support clinical decision-making during liver procurement. As part of developing a new spectrometry-based product, the company sought to complement the hardware with a digital solution capable of quantifying steatosis levels quickly and objectively. The system would allow surgeons to assess extracted livers without resorting to invasive methods like biopsies or frozen-section histology.

The Client turned to ScienceSoft to design a predictive software component to integrate with the spectrometer device. The goal was to train a model capable of estimating steatosis percentage based on spectrometry scan data. The Client envisioned a solution that would function across hospital settings, empowering procurement surgeons to perform and review scans via a mobile app, while administrators and analysts would manage the platform through web interfaces.

To succeed, the new system needed to operate reliably in high-stakes, time-sensitive environments with varying internet availability. It also had to support geographically distributed teams, provide consistent data capture across hospital sites, and ensure traceability for audits and clinical validation. On top of that, the Client aimed to establish a robust foundation for continuous improvement of its prediction models through centralized data collection and training.

AWS-Based Predictive Platform for Liver Graft Assessment

The project began with a discovery phase led by ScienceSoft’s business analyst and solution architect. Working closely with the Client, the team captured detailed functional and non-functional requirements for all user roles involved in the liver assessment workflow. This included surgeons at different levels of seniority, hospital-level administrators, and global platform managers.

To ensure clarity in UX expectations and future development, ScienceSoft created comprehensive storyboards, user flow diagrams, and functional maps for each system module. These assets helped align the team on role-specific needs and supported smooth communication with the Client’s subject matter experts throughout implementation.

Based on the discovery findings, ScienceSoft proposed a modular, AWS-based platform architecture designed to meet the project’s scalability, availability, and data privacy requirements, including compliance with HIPAA and GDPR. At its core, the platform used a neural-network-based regression model for steatosis prediction. ScienceSoft selected Apache PredictionIO as the backbone for the model orchestration, leveraging Spark for distributed data processing, HBase for event storage, and HDFS for storing training datasets and models. Model metadata, performance metrics, and evaluation results were stored in Elasticsearch to support monitoring and iteration.

All data exchange across the system was encrypted and handled through secure HTTPS connections and JSON APIs. Real-time workflow notifications were delivered via push mechanisms to ensure prompt updates for medical staff.

The platform’s predictive capabilities are made available through three tightly integrated components:

• An iOS mobile app serves as the primary tool for data collection during liver procurement. Junior surgeons use the app to connect to a spectrometer device, capture liver tissue scans, and submit results for senior review. The system then notifies the assigned senior surgeon, who evaluates the scan and makes the final procurement decision. The app supports offline operations and synchronizes data once online.
• A web application supports the review of liver scan data, procurement decision-making, and surgical team management. Junior and senior surgeons can view scans relevant to their teams, filter data, export records, and manage team membership. Both surgeons and hospital-level administrators have access to Power BI dashboards, with role-specific metrics: surgeons see scan outcomes, donor information, and team activity, while administrators monitor platform-wide usage trends, team composition, and performance indicators.
• An administrative panel allows global admins to manage platform settings, assign roles across hospitals, and oversee system usage. It features a graphical interface for managing the predictive model, including launching training cycles and evaluating new datasets. Dedicated reporting dashboards present key metrics such as scan volumes, general prediction accuracy, and admin activity across all connected facilities.

ScienceSoft delivered the platform in several phases, beginning with an MVP that included the core mobile workflows, role management, initial Power BI dashboards, and the first iteration of the predictive model. Subsequent releases added user feedback collection, GPS-based scan tracking, and integration with a national healthcare registry.

Better Collaboration, Decision-Making, and Traceability in Liver Transplant Workflows

The Client received a cloud-based software platform that supports real-time liver assessment during organ procurement. The solution enables structured data capture, remote collaboration among surgical team members, and objective decision-making without invasive procedures. Its modular AWS-based architecture ensures scalability, integration flexibility, and secure data handling in compliance with HIPAA and GDPR requirements.

The delivered system positions the Client to evolve its product ecosystem and explore new opportunities in the digital transplant tools market, complementing its medical hardware products with intelligent, decision-support software.

Technologies and Tools

AWS, Amazon S3, Amazon CloudFront, Elastic Load Balancer (ELB), AWS EC2, EC2 Security Groups, Apache PredictionIO, Apache Spark, Apache HBase, HDFS, Elasticsearch, MongoDB, PostgreSQL, Microsoft Power BI, Swift.