US20260268481
2026-09-10
Physics
G06T7/0012
The patent describes an advanced system that utilizes Artificial Intelligence (AI) and Machine Learning (ML) to detect health abnormalities by analyzing a patient's excreta. This system is integrated into a toilet equipped with multiple sensor systems. These sensors are designed to detect the user's presence, identify them, and capture relevant data during bathroom sessions. The collected data includes physical, chemical, and biological characteristics of the user's effluent, such as composition, color, and consistency. AI algorithms then process this data to draw conclusions about the user's health based on the effluent's constituents and usage frequency.
The system addresses the need for accurate, convenient, and unbiased electronic biomonitoring capabilities in bathroom settings. Historically, visual examination of human excreta has provided insights into health, but self-reporting is often subjective and prone to bias. Traditional bathroom components like toilets and mirrors have remained largely unchanged for over a century, lacking modern sensor or network capabilities. The invention aims to fill this gap by providing a method to objectively assess excreta, thereby aiding in tracking health symptoms and the effects of various interventions such as medications and diet.
The system comprises hardware and software components, including sensors placed strategically on the toilet to detect effluent characteristics. The sensors communicate data to an AI system, which is trained using ML algorithms to identify user behavior and health indicators by comparing new data with existing models. The system can also capture images of the effluent and analyze them against a database to diagnose health conditions. The invention is designed to be flexible, capable of being implemented in various configurations of hardware and software.
In a clinical study lasting over 12 months, TL Sensors were used to collect data, creating a "gold-standard" dataset validated by gastroenterology experts. The dataset involved digital biomarkers for excreta and toileting events, which were used to train ML models. These models can extract information about stool and urine characteristics, such as consistency, color, and frequency. The study collected over 2 million images, with 40,000 labeled images corresponding to 10,000 toileting sessions and over 100,000 digital biomarkers.
The TL Sensor system demonstrated superior accuracy compared to human caregivers in an assisted living community. It recorded significantly more bowel movements and urination events than staff, highlighting the limitations of manual documentation. The system reliably captured important characteristics like cloudy urine and loose stools, which are often missed by caregivers. The electronic monitoring capabilities of the TL Sensor were associated with detecting conditions such as urinary tract infections, dehydration, and other health issues, showcasing its potential impact on healthcare monitoring and diagnostics.