The FDA guidance document on when to submit a software change include this definition:
“Core algorithm” changes are modifications made to an algorithm that directly impact or contribute to the device’s intended use. Examples include: alarm algorithms on a monitor, a motor control algorithm for an infusion pump, and a detection module and measurement engine algorithm for an IVD.
Internally there is some back and forth on what constitutes an algorithm in the product. So I'm coming here for regulatory opinion.
Our device is a PACS (picture archiving and communications system) that is use by radiology for displaying scans to look for lesions suspicious of cancer, segment those lesions (draw the contours of it in 2D and display and calculate information about it as a 3D object), and provide data used for diagnoses and treatment. The radiologist draws what is thought to be the longest axis of the lesion and our device finds the contours and builds the 3D structure of the lesion across the scan "slices." Some people refer to the code that finds the extent of the lesion in 2D and 3D as our algorithm. Sometimes it's even referred to as our core algorithm (but without knowledge of the definition above).
Our intended use statement includes this sentence: The product automatically and semi-automatically segments normal structures and abnormal structures (for example, nodules and lesions), and provides metrics for the structures. (And in our last submission we defined the code/algorithm that does this activity.)
Others look at the FDA definition and say that's only for "black box" type of code like artificial intelligence software (both static AI and AI that modulates based on input).
Does anyone have experience internally or with the FDA when they have defined what they consider to be a core algorithm?
“Core algorithm” changes are modifications made to an algorithm that directly impact or contribute to the device’s intended use. Examples include: alarm algorithms on a monitor, a motor control algorithm for an infusion pump, and a detection module and measurement engine algorithm for an IVD.
Internally there is some back and forth on what constitutes an algorithm in the product. So I'm coming here for regulatory opinion.
Our device is a PACS (picture archiving and communications system) that is use by radiology for displaying scans to look for lesions suspicious of cancer, segment those lesions (draw the contours of it in 2D and display and calculate information about it as a 3D object), and provide data used for diagnoses and treatment. The radiologist draws what is thought to be the longest axis of the lesion and our device finds the contours and builds the 3D structure of the lesion across the scan "slices." Some people refer to the code that finds the extent of the lesion in 2D and 3D as our algorithm. Sometimes it's even referred to as our core algorithm (but without knowledge of the definition above).
Our intended use statement includes this sentence: The product automatically and semi-automatically segments normal structures and abnormal structures (for example, nodules and lesions), and provides metrics for the structures. (And in our last submission we defined the code/algorithm that does this activity.)
Others look at the FDA definition and say that's only for "black box" type of code like artificial intelligence software (both static AI and AI that modulates based on input).
Does anyone have experience internally or with the FDA when they have defined what they consider to be a core algorithm?