Teleradiology MVP
PROJECT TYPE: DATA DRIVEN SOFTWARE DEVELOPMENT
​
Tools and Programming Language: Power BI, Microsoft Power Apps, Microsoft SQL Server, PHP, HTML/CSS/JS, Data Studio, DCM4CHEE, Docker, Postgre SQL, Oviyam DICOM Viewer
​
BACKGROUND/ CONTEXT
​
Mid 2019, the firm I worked carried out of survey of the radiology space in Nigeria. We reached out to over 1000+ medical facilities and 100+ radiologists. In speaking with both facilities and radiologists, we came to an understanding of a service that could address pain points via insights derived from the data.
​
Late 2019, the firm decided to build a teleradiology platform. Via referrals, we reached out to Finnish software giants - Riftal. Competent at what they do, however with little to no expertise in health software there was a lot of back and forth with little progress. 3 months in, the calls and meeting petered out.
​
Early 2020, the IT manager and I decided to tackle the project rather than outsource it. Combined, we boasted over 4 years of deploying, managing, troubleshooting and resolving software/platforms in the teleradiology space. It seemed very logical to be able to take on development. The goal was to develop a minimum viable product (MVP), create a features roadmap for 12-18 months, and then onboard a software development firm to continuously iterate on that.
​
PROCESS
​
This project is divided into two phases, the data gathering and analysis phase, the MVP development phase.
​
Data Gathering/ Analysis
In reaching out to the medical facilities and radiologists, there was a two-sided constraint we had to work with - The data had to be stored centrally. On one side, the data should only be accessed while in the office and on the other side, anyone with the proper credentials should be able to add to the data remotely if needed. Several of the radiologists granted interviews outside of office hours and we had to accommodate such preferences.
To work around this constraint, I set up a Microsoft SQLServer on one for the inhouse servers. The data could be accessed, updated via a simple webpage built with HTML, CSS and PHP. For the other side of the constraint, I developed an app using Microsoft Power App and connected it to the SQLServer via Microsoft OnPrem Gateway. On completing this, phase one glided to a conclusion real quick.
​
MVP Development
​
In analyzing the data, several insights were discovered. I built a Power Bi report to share with the team and spark discussions. A few brainstorming sessions followed and we came up with a service and the minimum set of features needed to get it up and running. I had to build the same report using Data Studio to enable discussions with Board/ possible Investors.
​
When the MVP phase commenced, we setup a Linux Server as our development environment. The main components were containerized in docker instances. This allowed for modularity. Each component could be swapped and replaced as new knowledge came to light. The world of health tech -especially the niche that interfaces with big machines that occupy a whole room - is pretty slow in catching up to the more recent and fancier software frameworks. Documentation is dated. Well documented components are often no longer supported. The most recent information is found on obscure software discussion groups and support forums. Every now and then, you come across a perfect solution, only to find out it doesn’t perform well on certain kinds of hardware. All this hurdles and more were crossed to find the combination of components that could deliver the minimum set of features we had proposed.
RESULT
​
The components tested was documented. We settled for the Java-based DCM4CHEE (and its supporting softwares) and Oviyam, PHP, HTML, CSS and a few JavaScript frameworks. The MVP was successfully tested inhouse. It was also tested by a few preselected radiologists. Leveraging common customer complaints from previous deployments, we also came up with a features-roadmap.