Adaptive Machine Learning for Localized Anatomical Landmark Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning algorithms for identifying anatomical landmarks in cardiac MRI are pre-trained generically and do not account for local preferences of clinicians at specific clinical sites, leading to their underutilization due to diversity in landmark definitions across sites.
Innovation Solution
A system that allows for local on-site retraining of pre-trained machine learning algorithms based on feedback from clinicians during medical procedures, enabling the algorithms to adapt to local preferences without explicit prompting, thus improving the accuracy of anatomical landmark identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pre-trained machine learning algorithms are centrally trained according to global clinician preferences, then the algorithms can be deployed across multiple clinical centers, but they do not account for local preferences of doctors at each clinical site
Solution Approach 1:
The system performs preliminary action by collecting and storing feedback data from clinicians during normal operations, preparing the dataset in advance for future retraining. This allows the system to quickly adapt to local preferences when retraining is initiated, without requiring extensive data collection at the time of deployment to each clinical center.
Solution Approach 2:
The system enables self-service by allowing each clinical site to automatically retrain the machine learning algorithm using locally collected feedback data. The retraining process is initiated and managed at each clinical site independently, allowing local preferences to be incorporated without requiring centralized intervention for each site's customization needs.
2Productivity
If conventional pre-trained machine learning algorithms are deployed at multiple clinical centers, then deployment efficiency is improved, but the algorithms are underutilized due to diversity in landmark definitions across sites
Solution Approach 1:
The system implements feedback by collecting clinician corrections and preferences regarding anatomical landmark identification during normal operations. This feedback is stored and used to retrain the machine learning algorithm, creating a closed-loop system that continuously improves accuracy based on actual user interactions and local preferences.
Solution Approach 2:
The system applies dynamics by making the machine learning algorithm adaptable and changeable over time. Instead of deploying a static pre-trained model, the system enables the algorithm to dynamically adjust to local preferences through retraining with feedback data, transforming it from a fixed deployment to an evolving solution.
3Stability of the object's composition
If pre-trained machine learning algorithms are updated only when centrally managed software versions are released, then system stability is maintained, but updates are slow and do not incorporate local preferences
Solution Approach 1:
The system applies segmentation by separating the machine learning algorithm into independent, site-specific instances. Each clinical center can maintain its own version of the algorithm and retrain it independently using local feedback data, without affecting other sites. This allows parallel development and rapid updates at each location while maintaining overall system stability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods are described for automatically identifying an anatomical landmark in a medical image according to local preferences associated with a particular clinical site. A medical image for performing a medical procedure is received. An anatomical landmark is identified in the medical image using a pre-trained machine learning algorithm. Feedback relating to the identified anatomical landmark is received from a user associated with a particular clinical site. The feedback is received during a normal workflow for performing the medical procedure. The pre-trained machine learning algorithm is retrained based on the received feedback such that the retrained machine learning algorithm is trained according to local preferences associated with the particular clinical site.