System for automated AI-supported detection of contamination on surgical instruments

Abstract

System for the automated detection of contamination on surgical instruments, comprising: a) a camera unit for digital image acquisition of the surface of the surgical instrument to be tested, b) a data processing unit on which a pre-trained, non-fine-tuned convolutional neural network (CNN) is executed as a feature extractor with fixed weights, a few-shot classification module based on cosine similarity comparisons, an HSV contamination detection module, and a PaDiM-based anomaly scoring module are executed, c) a persistent reference database for storing class-specific model packages, comprising feature vectors, prototype vectors, and statistical distribution parameters, d) an output unit for displaying the binary test status (CLEAN / CONTAMINATED), the confidence values, and the test protocols with timestamps, wherein the system is configuredto transform an image captured by the camera unit into a feature vector using the CNN, to determine the instrument type by cosine similarity comparison against the reference database, to analyze the image in the HSV color space for biological contamination, to additionally perform a PaDiM-based anomaly scoring and to output a binary test status with a digital test protocol.