Patient-Specific Feature Ranking for Precision Diagnostics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Diagnostic systems based on machine learning algorithms provide population-wide rankings of clinical features, which may not be optimal for individual patients, leading to undesirable outcomes, especially in emergency situations, as they do not account for patient-specific conditions.
Innovation Solution
A method and system that rank unmeasured features for a patient by imputing values while holding other features constant, evaluating outcomes with a model, and determining statistical parameters to assign a ranking, allowing for the selection of a filtered dataset based on measured features from a master dataset, optimizing feature collection for accurate and timely predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If population-wide feature rankings are used for all patients, then the system complexity is reduced and data collection is simplified, but the diagnostic accuracy and relevance for individual patients deteriorates
Solution Approach 1:
The patent applies local quality by transitioning from uniform population-wide feature rankings to patient-specific feature rankings. The system evaluates and ranks features individually for each patient based on their unique characteristics, ensuring that the most relevant features are selected for each specific case rather than applying a one-size-fits-all approach.
Solution Approach 2:
The patent implements dynamics by making the feature ranking adaptive and changeable based on patient-specific inputs. The system dynamically adjusts feature rankings according to individual patient data, allowing the diagnostic approach to evolve and optimize for each patient rather than remaining static across all patients.
2Loss of information
If comprehensive feature sets are collected for all patients, then the potential diagnostic information is maximized, but the time and cost for data collection increases
Solution Approach 1:
The patent applies the taking out principle by extracting and selecting only the most relevant features for each individual patient from the comprehensive feature set. Rather than collecting all possible features for every patient, the system identifies and extracts the specific subset of features that are most diagnostic for that patient's condition, reducing unnecessary data collection.
Solution Approach 2:
The patent implements partial action by collecting only the necessary portion of features required for accurate diagnosis of each patient. The system determines the optimal number and type of features to collect based on patient-specific needs, avoiding the excessive collection of all possible features when fewer would suffice for the diagnostic task at hand.
3Measurement precision
If more features are measured for each patient, then the predictive accuracy improves, but the cost and time for data collection increases
Solution Approach 1:
The patent applies parameter changes by adjusting the number and type of features measured based on patient-specific parameters and conditions. The system modifies the feature collection parameters dynamically, selecting which features to measure and how many to collect based on the individual patient's characteristics, rather than using a fixed measurement protocol for all patients.
Data Source
AI summary
A method for ranking an unmeasured feature for an instance given at least one feature is measured is provided. The method includes imputing a first value to the unmeasured feature in the instance while holding the other remaining unmeasured features constant and evaluating a first outcome with a model using the first value in the instance. The method includes imputing a second value to the unmeasured feature in the dataset while holding the other remaining unmeasured features constant, evaluating a second outcome with the model using the second value in the instance, and determining a statistical parameter with the first outcome and the second outcome. The method also includes assigning the unmeasured feature a ranking corresponding to the determined statistical parameter. A system and a non-transitory, computer readable medium storing instructions to perform the above method are also presented.