⚛️ quantum physics

Hydration Monitoring Using Urinary Biomarkers: A Hybrid Classical Quantum Predictive Modeling Framework

This paper proposes a hybrid classical-quantum predictive modeling framework, featuring a modular Quantum Sequential Model, to monitor hydration status using urinary biomarkers from the Predict Health Toilet system, demonstrating the potential and current limitations of near-term quantum machine learning in digital health applications.

Original authors: Saul Gonzalez-Bermejo, Tommaso Albrigi, Borja Vazquez-Morado, Urko Regueiro-Ramos, Daniel Casado-Fauli, Sergi Consul-Pacareu, Laia Alentorn, Jordi Ferre, Valentino Asole, Parfait Atchade-Adelomou

Published 2026-04-20

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Saul Gonzalez-Bermejo, Tommaso Albrigi, Borja Vazquez-Morado, Urko Regueiro-Ramos, Daniel Casado-Fauli, Sergi Consul-Pacareu, Laia Alentorn, Jordi Ferre, Valentino Asole, Parfait Atchade-Adelomou

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine your body is a high-tech city. To keep the lights on and the traffic flowing, the city needs a perfect balance of water. If the water supply gets too low (dehydration), the power plants (cells) struggle, the roads get clogged (blood thickens), and the city's computer (your brain) starts making mistakes.

Usually, checking the city's water levels is a hassle. You have to go to a lab, get a blood test, or use a dipstick in a cup. It's reactive, meaning you only know there's a problem after you feel thirsty or sick.

This paper introduces a new way to monitor the city's water supply: The "Smart Toilet."

Here is the story of how they tried to predict hydration using a toilet, some math, and a little bit of futuristic "quantum" magic.

1. The Smart Toilet: A Passive Detective

The researchers used a system called the Predict Health Toilet (PHT). Think of this toilet not as a porcelain bowl, but as a super-smart detective that sits in your bathroom.

Every time you use it, the toilet doesn't just flush; it takes a quick "snapshot" of your urine. It measures three key things:

How "thick" the urine is (Specific Gravity).
How many electrical signals it carries (Conductivity).
How much of it there is (Volume).

Because this happens automatically while you do your daily business, it's "passive monitoring." You don't have to remember to take a test; the toilet does it for you.

2. The Problem: Too Much Data, Not Enough Clarity

The toilet collects a massive amount of data. But raw data is like a pile of unsorted LEGO bricks. Just having the bricks doesn't tell you what the castle looks like.

The researchers needed a way to turn those raw numbers into a clear prediction: "Is this person dehydrated?" or "How much water do they need?"

To do this, they built Predictive Models. Think of these models as two different types of chefs trying to guess the recipe of a soup just by tasting a spoonful.

3. Chef #1: The Classical Machine Learning Chef

First, they used Classical Machine Learning.

The Analogy: Imagine a very experienced, traditional chef who has cooked a million soups. They know the rules: "If the soup is salty and hot, it needs water." They use powerful, standard computers (like the ones in your laptop or the cloud) to crunch the numbers.
The Result: This chef was excellent. They could predict hydration levels with high accuracy. They are the reliable, workhorse solution we use today.

4. Chef #2: The Quantum Machine Learning Chef

Then, they tried something new and futuristic: Quantum Machine Learning (QML).

The Analogy: Imagine a chef from the future who cooks in a "quantum kitchen." In this kitchen, ingredients can be in two places at once (superposition) and can talk to each other instantly across the room (entanglement). This chef uses a Quantum Computer (or a simulator of one) to find patterns that the traditional chef might miss.
The Challenge: Quantum computers are currently very fragile and small (like a kitchen with only two burners). You can't cook a huge feast on them yet.

5. The Two Quantum Recipes

The researchers tried two different ways to cook with this quantum kitchen:

Recipe A: The "Symmetry" Chef (SU Models)
- This chef follows strict rules. They only use specific, symmetrical movements.
- Pros: Very stable and easy to understand.
- Cons: Because they follow strict rules, they are a bit rigid. In the experiment, this chef didn't cook as well as the traditional one. They were too limited by the small size of the quantum kitchen.
Recipe B: The "Modular" Chef (QSM)
- This chef built a flexible, Lego-like system called the Quantum Sequential Model (QSM). They could swap out parts, add layers, and re-upload data to make the dish more complex without needing a massive kitchen.
- Pros: Very flexible and adaptable.
- Cons: It's still learning. In the experiment, this chef did a decent job (almost as good as the traditional chef), but not quite perfect yet.

6. The Big Reveal: Who Won?

The researchers compared the results:

The Traditional Chef (Classical AI) won the cooking contest. They were the most accurate and reliable.
The Modular Quantum Chef (QSM) came in second. They showed great promise and proved that quantum computers can do this kind of work, but they aren't quite ready to replace the traditional chef yet.
The Strict Symmetry Chef came in third. They were too rigid for this specific task.

Why Does This Matter?

You might be thinking, "If the traditional chef won, why bother with the quantum one?"

Here is the metaphor: The Quantum Kitchen is still under construction.
Right now, quantum computers are like a prototype car engine. It's not faster than a V8 engine yet, but it runs on a completely different principle. If we can figure out how to make it work perfectly, it might one day solve problems that are impossible for traditional computers—like spotting complex disease patterns years before they happen.

The Takeaway

This paper is a "proof of concept." It says:

Smart Toilets work: We can monitor health passively and accurately.
AI works: We can predict hydration from toilet data.
Quantum AI is possible: We can run these predictions on quantum computers, even if they aren't perfect yet.

The future goal is to combine the Smart Toilet with the Quantum Chef. One day, your toilet might not just tell you to drink water; it might use quantum magic to warn you about kidney issues or metabolic changes before you even feel sick, all while you're doing your morning business.

In short: They built a smart toilet, taught it to predict thirst using old-school math, and then tried to teach it using futuristic quantum math. The old-school math won today, but the quantum math is showing it has a bright future.

1. Problem Statement

The study addresses the challenge of continuous, passive hydration monitoring in everyday environments. Traditional methods (e.g., plasma osmolality tests, urine strips) are reactive, episodic, and require active clinical intervention, often missing early signs of dehydration.

Goal: To develop predictive models that estimate hydration status using non-invasive, continuous data collected from smart sanitation systems.
Data Source: The Predict Health Toilet (PHT) system, which captures urinary biomarkers (Urine Specific Gravity, Conductivity, and Volume) during routine use.
Core Question: Can Quantum Machine Learning (QML) architectures, specifically hybrid quantum-classical models, offer advantages over classical machine learning in modeling physiological biomarker data, particularly given the constraints of current Noisy Intermediate-Scale Quantum (NISQ) hardware?

2. Methodology

A. Data Preprocessing and Feature Engineering

Input Features: A vector $X_t$ containing Urine Specific Gravity ( $USG_t$ ), Conductivity ( $C_t$ ), and Volume ( $V_t$ ).
Standardization: All features undergo z-score normalization to ensure numerical stability.
Dimensionality Reduction: Principal Component Analysis (PCA) is applied to reduce feature dimensionality. This is critical for QML models to fit within the limited qubit counts of current NISQ hardware while preserving 95–99% of the dataset's variance.

B. Modeling Frameworks

The study compares three distinct modeling approaches:

Classical Baselines:
- Model: XGBRegressor (Gradient Boosting).
- Role: Serves as the performance benchmark, optimized via Hyperopt for hyperparameters.
Symmetry-Constrained Quantum Regressors (QML-SU):
- Architecture: Utilizes unitary operators restricted to specific symmetry groups (e.g., $SU(3)$ on 2 qubits).
- Mechanism: Imposes strong inductive biases by restricting the accessible dynamical subspace.
- Goal: To test if architectural constraints improve trainability and interpretability on small-scale quantum circuits.
Quantum Sequential Model (QSM):
- Architecture: A modular, layer-wise compositional framework inspired by deep learning.
- Components:
  - Encoding: Angle embedding ( $R_Y$ rotations).
  - Variational Blocks: Trainable layers with entangling operators (e.g., CNOT rings).
  - Data Re-uploading: Injecting input data multiple times across layers to increase expressivity without deep circuits.
  - Measurement: Observable expectation values fed into a classical "head" (linear or shallow neural network) for final regression.
- Goal: To provide a flexible, scalable hybrid pipeline capable of exploring larger hypothesis spaces.

C. Experimental Setup

Infrastructure: Hybrid environment using Microsoft Azure ML for classical models and the UniQuE platform for quantum experiments (simulated on quantum circuit simulators).
Metrics: $R^2$ (Explained Variance), MAE (Mean Absolute Error), and Clinical Tolerance Accuracy ( $Acc_\delta$ ).

3. Key Contributions

Novel Hybrid Framework: Introduction of the Quantum Sequential Model (QSM), a modular architecture designed to bridge classical deep learning concepts with variational quantum circuits for physiological data.
Empirical Evaluation of QML in Digital Health: One of the first studies to systematically compare classical, symmetry-constrained, and flexible QML models on real-world urinary biomarker data.
Integration with Smart Infrastructure: Demonstrates the feasibility of using the Predict Health Toilet (PHT) system for generating structured, longitudinal datasets suitable for advanced predictive analytics.
Architectural Analysis: Provides a comparative analysis of how inductive biases (via symmetry constraints) vs. flexibility (via QSM) affect performance under NISQ constraints.

4. Experimental Results

The results were evaluated across three prediction tasks: Urine Specific Gravity (USG), Conductivity, and Volume.

Model Type	USG Prediction ( $R^2$ )	Conductivity ( $R^2$ )	Volume ( $R^2$ )	Key Observation
Classical (XGB)	0.91	0.88	0.90	Consistently achieved the highest accuracy and lowest error.
QML-SU (Symmetry)	0.52	0.49	N/A	Showed stable learning but significantly lower performance; failed on Volume prediction.
QSM (Flexible)	0.84	0.67	0.89	Performed competitively, approaching classical results on USG and Volume, demonstrating the value of flexible architectures.

Performance Gap: Classical models outperformed all quantum models. However, the QSM significantly outperformed the symmetry-constrained models, suggesting that flexibility is more critical than rigid symmetry for this specific dataset.
Error Analysis: Residual analysis indicated that while quantum models captured general trends, they struggled with high-variance outliers compared to classical ensembles.
Hardware Context: All quantum experiments were conducted on simulators. The authors note that current NISQ noise and limited circuit depth likely constrain the quantum models' ability to fully leverage quantum advantages.

5. Significance and Future Directions

Digital Health Potential: The study validates the concept of "smart sanitation" as a viable platform for non-invasive, continuous health monitoring, potentially enabling early detection of dehydration, renal dysfunction, and metabolic issues.
Role of Quantum Computing: While QML did not surpass classical methods in this specific instance (likely due to simulation constraints and data complexity), the QSM architecture proved viable. It offers a pathway for future hybrid systems where quantum processors may handle specific feature representation tasks that are difficult for classical computers.
Future Work:
- Deploying these models on real quantum hardware to test for intrinsic quantum advantages (entanglement/interference).
- Scaling to larger, more diverse physiological datasets.
- Extending the framework to detect broader metabolic and systemic health conditions beyond hydration.

Conclusion: The paper establishes a robust baseline for hybrid quantum-classical modeling in digital health. It concludes that while classical methods currently dominate in accuracy, flexible quantum architectures like the QSM are promising candidates for future integration as quantum hardware matures.