Partial Pill Recognition via Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Accurately determining the amount of medication absorbed by a person who has regurgitated after swallowing pills is challenging due to weight changes from moisture absorption and potential pill disintegration, making it difficult to assess the undigested portion.
Innovation Solution
A system and method using a recognition model to analyze images of regurgitated stomach contents, identifying pills and calculating the percentage of undigested medication, which can trigger specific actions based on medical data and the identification process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If weighing a regurgitated pill is used to determine absorbed medication, then the measurement process is simple, but the measurement precision deteriorates due to moisture absorption and weight changes
Solution Approach 1:
The patent replaces the mechanical weighing system with an image processing system using computer vision and machine learning models. Instead of measuring physical weight, the system captures images of regurgitated pills and uses algorithms to estimate the undigested portion, thereby avoiding moisture interference while maintaining measurement capability
Solution Approach 2:
The patent creates a visual copy (image) of the regurgitated pill and analyzes this copy to determine the undigested portion. The image serves as a substitute for direct physical measurement, allowing non-contact assessment that is not affected by moisture absorption or weight changes
2Measurement precision
If waiting for the pill to dry is used before measurement, then the moisture interference is eliminated, but the loss of time increases significantly
Solution Approach 1:
The patent replaces the time-consuming physical drying process with an immediate image processing system. The computer vision model analyzes the pill in its current state (wet or dry) and compensates for moisture effects through image analysis, eliminating the waiting period entirely
Solution Approach 2:
The patent performs the measurement action immediately upon capturing the image, without requiring preliminary drying. The image processing system is designed to handle wet surfaces and extract meaningful information about the undigested portion directly from the captured image
3Device complexity
If traditional image analysis is used to identify pills, then the device complexity remains low, but the difficulty of detecting and measuring increases due to partial digestion and disintegration
Solution Approach 1:
The patent replaces traditional image analysis with advanced machine learning models including convolutional neural networks (CNNs). These models are specifically trained to recognize pill characteristics even when partially digested, disintegrated, or obscured, significantly improving detection accuracy while maintaining reasonable system complexity
Solution Approach 2:
The patent transforms the image data into multiple feature parameters that the machine learning model can analyze. By extracting and analyzing multiple characteristics (shape, texture, color, edges) simultaneously, the system achieves high accuracy in identifying partially digested pills without requiring overly complex hardware
Data Source
AI summary
Methods, systems and computer program products for recognition of partially digested medications are provided. Aspects include receiving an image depicting regurgitated stomach contents of an individual and obtaining medical data regarding the individual. Aspects also include analyzing the image, by a recognition model, to identify one or more pills depicted in the image and a percentage of the one or more pills that has not been digested. Aspects further include performing an action based on the medical data, the identification of the one or more pill and the percentage of the one or more pills that has not been digested.