Eye Imaging SaMD Debugged and Market-Ready in 6 Weeks
About Our Customer
The Customer is a US-based provider of advanced diagnostic and surgical devices. It equips research institutions and healthcare providers worldwide with innovative confocal scanning systems that enable the quantitative measurement of living tissue biomechanics.
Eye Imaging System Had UI Flaws and Lacked Precision
The Customer was preparing to release an eye imaging solution, which consisted of an optical scanner and a desktop SaMD (Software as a Medical Device) application. The solution examined eye tissue stiffness by measuring the wavelengths of scattered laser light and conducting real-time spectral analysis. However, the app had numerous bugs, which could affect user experience and hinder the solution’s performance. The Customer was also looking into improving the accuracy of the system’s heatmap generation algorithm.
The company was searching for a dependable medical technology vendor to fix the existing bugs and check the viability of the current and potential imaging algorithms. The Customer chose ScienceSoft, trusting our experience in healthcare IT and proficiency in image analysis. We promptly assembled a team that was to complete the project within six weeks due to the tight deadline set by the Customer.
Resolving UI Defects to Enhance User Experience
The desktop SaMD app comprised several integrated modules written in Java and C++.
The app’s user interface was developed on JavaFX, providing an interactive and visually appealing view of the ocular biomechanics findings. The Spring Framework was used for all the back-end operations, including database management, data manipulation, and system orchestration.
ScienceSoft’s team tested the software modules for defects and revealed that the overwhelming majority of the detected bugs were related to faulty UI elements, for example:
- Scrollbars didn’t work, leading to problems with navigating large data sets.
- Window size couldn’t be adjusted, causing the graphs and tables used for data analysis to be cut off or distorted.
- Buttons and checkboxes weren’t responsive, leaving users in doubt about whether their actions were registered or not.
- Color contrast wasn’t sufficient, making it difficult to find and interpret critical ocular biomechanics data.
Our team promptly fixed the discovered defects, ensuring seamless functionality for the end users.
Improving Heatmap Accuracy
The Customer considered replacing the existing kernel destiny estimation (KDE) algorithm with Gaussian processes (GP). Being nonparametric, the KDE method estimates the density of measurements at various points in the eye tissue and creates a spatially resolved profile of ocular tissue stiffness. GP is a probabilistic algorithm that models the underlying spatial distribution of data points and generates interpolated values at each point.
To evaluate the feasibility of shifting from KDE to GP, ScienceSoft compared the heatmaps generated by both algorithms with manually labeled stress regions provided by the domain experts. After careful analysis, we concluded that the KDE algorithm performed better despite some inaccuracies in detecting and depicting high-stress regions. Also, GP demonstrated high computational demands and limitations for real-time analysis on a desktop application.
Our team took a series of steps to enhance the KDE algorithm:
- Gathering additional data on stress patterns in the eye from experimental studies and computational simulations shared by the domain experts.
- Comparing the stress patterns generated by the algorithm with the additional data to identify discrepancies.
- Incorporating the info on new stress patterns into the algorithm's training process to estimate tissue stiffness and depict stress patterns more accurately.
- Calibrating the algorithm so that it can assign appropriate significance to different aspects of the stress patterns.
Innovative Eye Imaging App Market-Ready Within 6 Weeks
In just six weeks, ScienceSoft perfected the Customer's SaMD app and ensured that the scanning solution was ready for market launch. We improved the UI and fine-tuned the heatmap generation algorithm for higher accuracy.
The Customer was fully satisfied with the quality of the final solution and benefitted from the fast time-to-market for its eye imaging product.
