Surgical Microscopy Annotated Data Acquisition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Acquiring high-quality data for training machine learning instruments in the medical field is challenging due to existing medical processes being primarily designed for patient treatment and cost reduction, rather than data collection, which can disrupt medical procedures.
Innovation Solution
A method utilizing surgical microscopy systems to record and annotate images, where the systems determine relevant criteria automatically or with user input, and store them in a database for training machine learning instruments, minimizing disruption to medical workflows by only requesting annotations when criteria match predefined sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data acquisition processes are added to surgical microscopy systems to collect training data for machine learning instruments, then the quantity and quality of training data improve, but the complexity of the medical process increases and may disrupt existing workflows
Solution Approach 1:
The patent combines data acquisition functions with the existing surgical microscopy system, integrating image recording and annotation capabilities into the workflow without requiring separate dedicated data collection equipment. The surgical microscopy system simultaneously performs its primary surgical function and collects training data through integrated software modules that operate during normal surgical procedures.
Solution Approach 2:
The surgical microscopy system is designed to serve multiple functions: it performs surgical imaging for patient treatment while simultaneously acquiring annotated data for machine learning training. The system's image recording capability is utilized for both immediate surgical documentation and long-term data collection, eliminating the need for separate data acquisition processes.
2Manufacturing precision
If annotations are requested from users during surgical procedures to create annotated training data, then the quality of annotated data improves, but the time required for the surgical process increases
Solution Approach 1:
The system performs preliminary actions by automatically determining criteria from recorded images and pre-selecting images that match desired criteria before requesting user annotation. This pre-filtering process ensures that users only need to annotate a small subset of images that are most suitable for training purposes, rather than annotating all captured images during surgery.
Solution Approach 2:
The system implements feedback mechanisms where automatically determined criteria are compared against desired criteria, and only when there is sufficient correspondence does the system request user confirmation for annotation. This feedback loop allows the system to intelligently identify and prioritize images that require annotation, minimizing user burden while maintaining data quality.
3Productivity
If automated determination of criteria is used to identify images for annotation, then the productivity of data acquisition improves, but the measurement precision of criteria determination may decrease compared to expert annotation
Solution Approach 1:
The system uses an intermediary approach where automated software determination of criteria serves as a preliminary filter, and user expertise acts as a confirmation layer. The automated system identifies potential candidates for annotation based on image analysis, then presents these to users for final verification, combining the speed of automation with the precision of human judgment.
Solution Approach 2:
The system applies partial automation by using automated criteria determination only for initial image selection and filtering, rather than attempting to fully automate the entire annotation process. This partial approach allows automated processing of large volumes of images while reserving human expertise for the critical decision points where precision is most important.
Data Source
AI summary
A method for acquiring annotated data with the aid of surgical microscopy systems comprises obtaining desired criteria which are intended to be satisfied by desired data to be annotated, and storing the set of desired criteria in a plurality of surgical microscopy systems. In each surgical microscopy system, images are then recorded and current criteria which correspond to the recorded images are determined. The current criteria are compared with the desired criteria. If the desired criteria sufficiently correspond to the current criteria, a confirmation is requested from a user as to whether said user would like to annotate data. If the user provides the confirmation, annotations for images are received from the user and stored together with the images.