Fine-Grained Serum HILN Detection With Adversarial Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated diagnostic systems face challenges in accurately and efficiently determining the presence and degree of hemolysis, icterus, and lipemia in serum or plasma portions of specimens due to label obstruction and computational complexity, leading to subjective manual inspections and potential human error.
Innovation Solution
A method using a deep semantic segmentation network with over 100 layers processes multi-spectral, multi-exposure image data from multiple viewpoints to classify serum or plasma portions into hemolytic, icteric, lipemic, and normal classes, with sub-classes, to determine the presence and degree of interferents, accounting for label obstruction through HDR imaging and a segmentation convolutional neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual inspection is used to determine HILN status, then subjective assessment can be performed, but labor intensity increases and human error occurs
Solution Approach 1:
The patent replaces manual visual inspection with an automated machine vision system that captures images of the serum or plasma portion and uses image processing algorithms to objectively determine HILN status. This substitution eliminates human labor intensity and subjective bias while maintaining or improving detection accuracy through consistent, repeatable measurements.
2Ease of operation
If automated machine vision inspection is used for pre-screening, then labor intensity is reduced, but computational complexity increases due to image processing requirements
Solution Approach 1:
The patent segments the image processing task into distinct stages: image capture from multiple viewpoints, preprocessing to handle label obstructions and lighting variations, feature extraction to identify relevant visual characteristics, and classification to determine HILN status. This segmentation reduces computational complexity by processing only essential features rather than analyzing entire images, while maintaining high automation levels.
3Measurement precision
If label obstruction is present on specimen containers, then automated inspection difficulty increases, but manual inspection remains subjective
Solution Approach 1:
The patent captures images from multiple viewpoints (front, back, and angled views) to overcome label obstruction. By examining the specimen from different dimensions and angles, the system can identify serum or plasma characteristics that are not obscured by labels, maintaining objective assessment accuracy while reducing inspection difficulty through multi-perspective analysis.
4Measurement precision
If fine-grained classification with sub-classes is implemented, then HILN detection precision is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary classification to identify the general HILN status first, then applies fine-grained sub-classification only to specimens that require more detailed assessment or show ambiguous characteristics. This staged approach maintains high classification precision for critical cases while reducing processing time for clear-cut specimens, optimizing the balance between accuracy and efficiency.
Data Source
Figure 1~2
Figure 3A~3B
Figure 4A~4B
AI summary
A method of characterizing a serum or plasma portion of a specimen in a specimen container provides a fine-grained HILN index (hemolysis, icterus, lipemia, normal) of the serum or plasma portion of the specimen, wherein the H, I, and L classes may each have five to seven sub-classes. The HILN index may also have one un-centrifuged class. Pixel data of an input image of the specimen container may be processed by a deep semantic segmentation network having, in some embodiments, more than 100 layers. A small front-end container segmentation network may be used to determine a container type and boundary, which may additionally be input to the deep semantic segmentation network. A discriminative network may be used to train the deep semantic segmentation network to generate a homogeneously structured output. Quality check modules and testing apparatus configured to carry out the method are also described, as are other aspects.