Building a robust system to track ovarian follicles

 

Building a robust system to track ovarian follicles

Background:

The therapists of the client’s organization were required to frequently pause ultrasound films in order to find and record the horizontal and vertical follicle dimensions. The discovered follicles had to be manually measured later. It took a long time to complete this process. Automatically detecting, segmenting, and measuring follicles from ultrasound scans was the client’s need.

The client prioritized detection accuracy over speed in order to overcome this obstacle. Accurate results and the detection of the most follicles were both crucial.

Building a robust system to track ovarian follicles

Solution:

Matching pixels with object sizes: From the client’s data, the consulting firm discovered that ultrasound photos include hatch markings that are spaced evenly. For accurate follicle measurement, the company developed algorithms for the precise measurement of the pixel size by detecting those hatch marks.

Researching and preparing specific data: The consultants had extended discussions with the client to get all the required information and to gain a thorough grasp of how doctors use this information. They outlined the specifications for annotating the data, and the client applied these specifications to create a high-fidelity dataset.

Building a robust system to track ovarian follicles<br />

Results:

They created a successful AI-based healthcare system that analyzes ultrasound video scans, finds follicles, and calculates the diameter, surface area, and perimeters of those follicles. Potential methods and difficulties for developing an AI solution for the given problems during the discovery phase were investigated. Based on the findings of the study, an initial accuracy of more than 70% was desired.

The initial results demonstrated 77% precision and 86% recall. With additional study, creation, and system training, the solution’s effectiveness was increased to 90% precision and 97% recall.

Building a robust system to track ovarian follicles<br />

Conclusion:

The client’s ability to identify and measure follicles was greatly aided by the given solution. By reducing the time required for doctors manually going through dozens of photos, they were able to review video files quickly, double-check the system’s findings, and focus more on patients.

Contact Us

What Happens Next?

  1. Our customer success manager will analyze your business requirements.
  2. We will sign an NDA to ensure your information security.
  3. Our pre-sale manager presents project estimates and approximate timeline