Probabilistic Finite State Automaton Models for Electronic Health Record Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing complex diseases such as MACE, Alzheimer's disease, Idiopathic pulmonary fibrosis, and autism spectrum disorder are invasive, expensive, and often ineffective, particularly in early stages, due to reliance on limited physical examinations and laboratory tests.
Innovation Solution
A computer-implemented method using machine learning algorithms to analyze electronic health records, generating probabilistic finite state automaton models to predict disease risk without requiring new tests or specific data, by partitioning records by gender and inferring sequence likelihood defects to identify potential health conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional physical examination and laboratory tests are used for diagnosing complex diseases, then diagnostic accuracy can be achieved, but the process becomes invasive, time-consuming, and expensive
Solution Approach 1:
The patent replaces traditional mechanical diagnostic methods (physical examinations, laboratory tests, imaging) with an information-processing system that uses machine learning algorithms to analyze electronic health records. The system substitutes direct physical measurement with computational analysis of historical data patterns, achieving diagnostic insights without invasive procedures or additional time consumption
Solution Approach 2:
The system creates a virtual model of patient health status by copying and analyzing patterns from historical electronic health records. Instead of performing new physical measurements, the system replicates diagnostic capabilities through computational models trained on existing data, eliminating the need for additional testing while maintaining diagnostic accuracy
2Productivity
If computer-assisted diagnosis is introduced to improve efficiency, then time consumption is reduced, but diagnostic accuracy remains limited by the same features available to physicians
Solution Approach 1:
The system fundamentally changes the parameters used for diagnosis by transitioning from traditional clinical features to thousands of derived features extracted from electronic health records using machine learning. This includes temporal patterns, diagnostic code sequences, and relationships between different health events that are invisible to human physicians but quantifiable by computational algorithms
Solution Approach 2:
The patent adds a new dimension to diagnostic analysis by incorporating temporal sequencing and pattern recognition across multiple time points. The system analyzes the sequence and timing of diagnostic codes, laboratory results, and clinical events over time, creating a multi-dimensional view of patient health trajectories that transcends static feature assessment
3Reliability
If existing risk calculators like RCRI are used to predict disease risk, then some risk assessment is provided, but many risk factors are not considered and patients without formal diagnoses are not identified
Solution Approach 1:
The system creates a universal risk prediction framework that can assess multiple disease risks simultaneously across diverse patient populations. The machine learning models are trained to recognize patterns associated with various complex diseases including cardiovascular events, Alzheimer's disease, pulmonary fibrosis, and autism spectrum disorder, providing versatile risk assessment that adapts to different clinical scenarios and patient characteristics
Solution Approach 2:
The system performs preliminary risk identification by analyzing electronic health records before clinical symptoms manifest or formal diagnoses are established. By detecting subtle patterns in historical data such as sequential diagnostic codes and laboratory trends, the system flags at-risk patients in advance, enabling early intervention before traditional diagnostic thresholds are met
Data Source
AI summary
A method including receiving a plurality of electronic health records stored in a database and partitioning the plurality of electronic health records in a first set of plurality of electronic health records and a second set of plurality of health records is disclosed. The method includes, for each electronic health record of the first and second sets of the plurality of electronic health records, generating a plurality of data streams, and in accordance with the generated data streams corresponding to the respective group of related disorders, inferring probabilistic finite state automaton (PFSA) models corresponding to a positive cohort and a control cohort for a specific health condition. The method includes determining a respective sequence likelihood defect of an electronic health record data of a new patient to match one of the inferred PFSA models for determining a likelihood of the new patient to acquire the specific health condition.