Bitmap-Based Sequential Pattern Mining for Healthcare Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sequential pattern mining techniques in healthcare fail to provide a clear understanding of the relationship between a patient's medical history and outcomes, primarily focusing on pattern counting rather than frequency analysis, and are computationally inefficient.
Innovation Solution
A computer-implemented method that constructs patient records into bitmap representations, analyzes these bitmaps to identify sequential patterns on a per-patient basis, and generates patient-based counts of pattern occurrences, enabling more accurate pattern detection and reducing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional sequential pattern mining techniques are used, then pattern counting can be performed, but computational efficiency is poor and per-patient analysis is not achieved
Solution Approach 1:
The patent transforms patient medical history data into a bitmap representation, changing the data structure from traditional tabular format to a binary matrix where rows represent patients and columns represent events. This parameter change enables efficient bitwise operations for pattern mining while preserving per-patient analysis capability, resolving the contradiction between computational efficiency and measurement precision.
Solution Approach 2:
The patent replaces traditional sequential pattern mining algorithms with a bitmap-based approach using bitwise operations. This substitution of the computational mechanism dramatically improves processing speed while maintaining the ability to identify sequential patterns at the patient level, addressing both computational efficiency and pattern detection accuracy.
2Productivity
If pattern counting is performed without per-patient analysis, then computational resources are reduced, but the ability to determine which patients experienced specific patterns is lost
Solution Approach 1:
The patent segments the analysis by organizing data into patient-specific rows in the bitmap structure. Each row represents a single patient's medical history, allowing the system to efficiently count patterns while preserving the ability to identify which specific patients experienced each pattern. This segmentation enables both computational efficiency and patient-level information retention.
Solution Approach 2:
The bitmap structure serves as an intermediary between raw medical history data and pattern analysis results. It efficiently encodes patient-specific information in a compact format that enables both resource-efficient processing and complete retention of patient-level pattern occurrence data, preventing information loss while improving computational efficiency.
3Ease of manufacture
If traditional sequential pattern mining is used, then implementation is simpler, but the ability to perform personalized prediction and feature selection is limited
Solution Approach 1:
The bitmap-based sequential pattern mining approach serves multiple functions: it performs traditional pattern counting, enables per-patient analysis, supports personalized prediction, and facilitates feature selection. This multi-functional implementation maintains relative simplicity while dramatically increasing adaptability for personalized healthcare applications.
Data Source
AI summary
Examples of techniques for patient-level analytics with sequential pattern mining are provided. In one example implementation according to aspects of the present description, a computer-implemented method includes: constructing a patient record; transforming, by a processing system, the patient record into a bitmap representation; and analyzing, by the processing system, the bitmap to identify a sequential pattern within the patient record on a per patient basis.