Margin Assessment Using Harmonic Generation Microscopy and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods struggle to accurately define surgical margins of Extramammary Paget's disease lesions, often leading to incomplete removal due to ill-defined tumor borders and extended tumor spread, which complicates visual assessment and treatment.
Innovation Solution
A margin assessment method combining non-linear harmonic generation microscopy with deep learning, where 3D image groups are generated, labeled, and used to train a deep learning model for accurate classification of malignant and normal skin tissue, enabling real-time, non-invasive margin determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional clinical features are used to assess skin cancer margins, then the assessment process is simple, but the accuracy is poor due to ill-defined tumor borders and extended tumor spread
Solution Approach 1:
The patent replaces conventional clinical visual assessment with non-linear harmonic generation microscopy (HGM), which uses optical fields to generate high-resolution 3D images of tissue structures. This substitution enables precise visualization of tumor boundaries and stromal changes that are invisible to clinical examination, achieving accurate margin assessment without requiring complex surgical exploration procedures
Solution Approach 2:
The patent transitions from 2D clinical surface examination to 3D volumetric imaging using HGM. By capturing multiple focal planes and generating three-dimensional reconstructions of tissue architecture, the system reveals the depth and spatial extent of tumor infiltration that cannot be detected in conventional two-dimensional clinical photos, thereby improving margin definition accuracy
2Reliability
If complete surgical removal is attempted with conventional methods, then treatment thoroughness is improved, but the difficulty in identifying accurate margins leads to incomplete removal
Solution Approach 1:
The patent introduces HGM imaging as an intermediary tool between surgical resection and pathological diagnosis. The 3D HGM images serve as a preoperative map that guides surgical margins, allowing surgeons to visualize and respect precise tumor boundaries before cutting, thereby reducing the difficulty of accurate margin identification and improving complete removal rates
Solution Approach 2:
The system performs preliminary 3D imaging and deep learning-based margin prediction before surgical resection. By pre-visualizing the complete extent of tumor infiltration and generating predicted margin maps, the system enables surgeons to plan and execute complete removal more effectively, transforming a postoperative diagnostic challenge into a preoperative planning advantage
3Measurement precision
If deep learning model is trained with labeled 3D image groups, then classification accuracy is improved, but the data processing and model training complexity increases
Solution Approach 1:
The patent creates digital copies of tissue structures through HGM imaging, generating 3D image groups that serve as training data. These digital representations capture the complete morphological information of both malignant and normal tissues, allowing the deep learning model to learn accurate classification patterns from high-fidelity digital twins of the tissue architecture
Solution Approach 2:
The system transforms conventional 2D histopathological images into 3D volumetric data through HGM, changing the dimensional parameter of the data representation. This transformation provides the deep learning model with richer spatial information and structural context, significantly improving classification accuracy while the automated processing pipeline manages the complexity of data preparation and model training
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach achieves high accuracy in distinguishing malignant EMPD from normal skin with sensitivity of 98.06%, specificity of 93.18%, and accuracy of 95.81%, facilitating precise margin assessment and complete lesion removal.
Implementation Method 1
extracting, by using a harmonic generation microscopy (HGM) imaging system, a plurality of 3D image groups within a range from a surface of a plurality of positions in the predetermined specimen region to a predetermined depth
Data Source
AI summary
A margin assessment method is provided. Under cooperation of harmonic generation microscopy (HGM) and a deep learning method, the margin assessment method can instantaneously and digitally determine whether a 3D image group generated by an HGM imaging system is a malignant tumor or the surrounding normal skin, so as to assist in determining margins of a lesion.