ML-powered Web Platform for Running Eye Lens Power Calculations

Client

An advanced laser eye surgery and comprehensive eye treatment clinic

Industry

Healthcare

Services

Product Development, ML, UX/UI, QA

Tech

Angular, AWS, REST API, SES, S3, RDS

Challenge

The client has developed a machine learning-driven formula for calculating the power of intraocular lens (IOL) implanted in patients as part of a treatment for eye conditions, such as cataracts or myopia. They wanted to promote and certify their proprietary innovative method by highlighting its accuracy among the world’s eye surgeon community that mostly uses four other formulas.

Specifically, our team took on the following challenges:

●

Create an easy-to-use web platform for running lens power calculations based on the four classic formulas and the client’s bespoke method

●

Implement a functionality for optimizing formula constants, which are used as adjustment factors to improve calculation precision based on the lens location and targeted postoperative refractive results

●

Develop an ML model for calculations using the client’s formula

●

Develop a functionality for building comparison charts of postoperative eye parameters relative to the target refraction for different formulas to understand their accuracy

Solution

●

A web platform, built as an Angular application with a cloud infrastructure on Amazon Web Services

●

The platform consists of three components: for optimizing formula constants, performing IOL power calculations and predicting refractive error

●

Attractive UI/UX ensuring intuitive and simple interaction with the system.

●

The platform allows eye care professionals from around the world to easily perform preoperative lens power calculations using the five formulas based on input data: lens types, actual eye measurements, the target refraction and eye status that can be “untreated eye” or “previous laser eyes”

●

Users can use default formula constants, or optimize constants clinical historical data for the four classic formulas depending on eye status and train the ML model for the client’s formula

●

Optimized constants are updated to the latest data upload and can be downloaded to an excel spreadsheet

●

The calculation results, i.e. the desired lens power, can be downloaded as a PDF file

●

A predictive algorithm delivers an easy-to-interpret and concise histogram of postoperative refraction results relative to the target refraction for different formulas to easily compare accuracy between them; a filter option is available to extract proper results, enabling users to select the lens type, eye axial length range, eye status (untreated or previous surgery), postoperative stage and corneal astigmatism range

Impact

●

The web platform provides a reliable and simple mechanism for optimizing constants and running the most accurate lens power calculations, ensuring optimal lens selection

●

By uploading historical data on eye surgeries, users can сonstantly train the client’s ML-driven formula to come up with a more robust and accurate prediction of effective lens position

●

The platform has helped the client to introduce its innovative formula into a wider usage among eye care professionals