Joint Cell Region Classification in Digital Pathology

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current digital pathology systems face challenges in efficiently processing and analyzing large images of tissue slides, particularly in joint cell detection, segmentation, and classification, which are crucial for accurate diagnosis and research.

Innovation Solution

The proposed solution involves a deep learning approach using a multilayer neural network that simultaneously classifies cells and regions within digital pathology images. This method leverages semi-automated image analysis and ground truth data generated by pathologists to improve classification accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional separate cell detection and region classification methods are used, then each task can be optimized independently, but the overall processing time and computational complexity increase significantly

Engineering Contradiction:
Improveclassification accuracyVSAvoidprocessing speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent merges cell detection and region classification into a single unified deep learning model that processes images at multiple scales simultaneously. The model integrates cell-level feature extraction with region-level semantic understanding in one architectural framework, enabling joint optimization of both tasks and eliminating the sequential processing overhead of separate methods.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces multi-scale processing by processing images at different resolution levels simultaneously. The deep learning model incorporates feature pyramids and multi-scale convolutional layers that analyze both local cell details and global region context in parallel, adding a dimensional aspect to the processing that enables simultaneous high-accuracy detection and classification.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If manual ground truth annotation by pathologists is performed, then classification accuracy is improved, but the time required for training data preparation increases

Engineering Contradiction:
Improveground truth accuracyVSAvoiddata preparation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent employs preliminary automated detection and segmentation algorithms that prepare initial annotations before pathologist review. These pre-processing steps include automatic cell detection, nuclei segmentation, and preliminary classification that create a draft ground truth dataset, reducing the time pathologists need to spend on manual annotation while maintaining high accuracy through expert verification.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If deep learning models are trained on large datasets, then classification performance improves, but the computational resources and training time required increase

Engineering Contradiction:
Improvemodel accuracyVSAvoidcomputational energy
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the training process into efficient stages using pre-trained models and transfer learning. Instead of training from scratch on large datasets, the system utilizes pre-trained deep learning models that have learned general features, then fine-tunes them on specialized pathology data. This segmentation of the training process significantly reduces computational energy requirements while maintaining high accuracy.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3721373B1Deep-learning systems and methods for joint cell and region classification in biological images
Publication Date: 2025.03.05 VENTANA MEDICAL SYSTEMS INC
  • EP3721373B1 patent drawingFigure 1
  • EP3721373B1 patent drawingFigure 2
  • EP3721373B1 patent drawingFigure 3A

AI summary

The present disclosure relates to automated systems and methods for training a multilayer neural network to jointly and simultaneously classify cells and regions from a set of training images. The present disclosure also relates to automated systems and methods for using a trained multilayer neural network to classify cells within an unlabeled image.