Follicle Clarity

Follicle Clarity Software is a cloud-based medical image analysis tool designed to automatically quantify the number and size of ovarian follicles from transvaginal ultrasound images using 'locked' machine learning algorithms and automatic segmentation. Its purpose is to aid clinicians in fertility assessment by providing objective and efficient follicle measurements, with its performance validated through comparative non-clinical and clinical studies.

Follicle Clarity is a cloud-based software as a medical device (SaMD) designed to assist clinicians in reproductive medicine. Its primary function is to automatically calculate the number and size of hypoechoic structures, specifically ovarian follicles, from transvaginal ultrasound images. The software receives DICOM-format ultrasound data, employs 'locked' (non-adaptive) machine learning algorithms and automatic segmentation technology to identify and measure follicles, and then displays this data both graphically and in tabular format. Users can input patient-specific information like cycle day, estradiol, and progesterone levels, and have the ability to review displayed ultrasound images and manually add or delete measurements, allowing for clinical oversight and adjustment of the automated results. Performance of Follicle Clarity was assessed through non-clinical and clinical studies. Non-clinical validation included machine learning algorithm training and phantom trials, which compared the software's measurements to manual measurements from a DICOM viewer, demonstrating accuracy within the margin of error for human measurements for various follicle densities. A prospective clinical study evaluated the software's accuracy, precision, and agreement against manual measurements by ultrasonographers and measurements by a predicate device (SonoAVC). The study's primary endpoint was the median size of ovarian follicles, and the secondary endpoint was the number of follicles identified; Follicle Clarity met these predetermined endpoints, showing comparable performance in identifying follicle number and size.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the 'locked' machine learning algorithm in Follicle Clarity compare to adaptive or continuously learning algorithms in terms of long-term accuracy and robustness across diverse patient populations and varying ultrasound image quality?; What specific pre-processing steps, if any, are applied to the transvaginal ultrasound images before the machine learning algorithm performs segmentation and quantification, and how sensitive is the algorithm to variations in these steps?; What were the exact statistical measures and thresholds used to define 'substantially equivalent performance' in the clinical validation study when comparing follicle count and size measurements against SonoAVC and manual methods?; How does Follicle Clarity handle complex cases, such as ovaries with a very high number of follicles (e.g., PCO-like) or images with artifacts, and what is the typical user intervention rate for correcting automated measurements?; Beyond follicle count and size, are there plans to expand Follicle Clarity's capabilities to include other relevant ovarian or endometrial features that could aid in fertility assessment, such as antral follicle classification or endometrial thickness/pattern analysis?