A Hybrid Fractional-Order Dynamical Model for ECG Morphology: Caputo Memory, Identifiability, and Reproducible PQRST Reconstruction

The human heart beats with a rhythm that is both steady and subtly variable, a pulse that doctors have long read by tracing the electrical waves it sends across the skin. This tracing, known as an electrocardiogram, or ECG, looks like a jagged line of hills and valleys, each peak and dip corresponding to a specific moment of the heart's electrical cycle. For decades, scientists have tried to build mathematical models that can recreate these waves, hoping to understand the heart's health or to simulate how it might behave under stress. Traditional models treat the heart like a simple machine that repeats the same pattern over and over, or like a complex system where every part remembers everything that happened before. However, finding the right balance between simplicity and accuracy has been difficult, often leaving researchers with models that either miss the subtle details of the heartbeat or become so complicated they are impossible to use.

A team of researchers from Ankara and Atılım universities has proposed a new way to look at this problem, one that treats the timing of the heartbeat and the shape of the wave as two separate but connected stories. In their work, they suggest that the heart's internal clock, which dictates when a beat happens, follows a standard, predictable rhythm. But the actual shape of the electrical wave—the way it rises, peaks, and falls—carries a kind of memory. This memory means that the current shape of the wave is influenced not just by the immediate moment, but by what happened slightly before, fading away slowly over time. By separating the clock from the shape, the researchers created a model that can reproduce the complex curves of a heartbeat with surprising precision, capturing details that older models often miss.

The researchers built a hybrid system to test this idea. They imagined the heart's phase, or its position in the beat cycle, as a simple, ordinary clock ticking forward. This clock drives a second part of the system, which is responsible for generating the actual voltage signal seen on the ECG. This second part uses a special type of mathematical rule that allows it to remember past events, but only for a short while. Think of it like a sponge that soaks up water; it holds onto the moisture for a time before slowly releasing it, rather than holding it forever or letting it drip away instantly. In their model, this "sponge" effect is applied only to the shape of the wave, not to the timing of the beat itself. This distinction is crucial because it avoids a mathematical trap where models fail to produce a repeating pattern if they try to apply this memory rule to the clock as well.

To see if this approach worked, the team created a synthetic, or computer-generated, heartbeat. They programmed this artificial heart to beat at a steady rate of 75 times per minute, with a specific shape for its electrical waves. They then added realistic imperfections to the signal, such as small random changes in the height of each beat, a slow drifting baseline that mimics the movement of a patient's chest, and a layer of static noise similar to what doctors see in real recordings. They then asked their new model to learn the rules of this artificial heart. The model had to figure out two main things: how strong the memory effect was, and how fast the signal should relax back to a resting state. It also had to estimate the size of the five main bumps that make up a heartbeat, known as the P, Q, R, S, and T waves.

The results of this test were clear. When the researchers compared their new model to a traditional one that did not include this memory effect, the new model was significantly better at predicting the shape of the heartbeats it had not seen before. In a specific test using four beats that were held back for validation, the new model made errors that were about 22 percent smaller than the traditional model. It also matched the overall pattern of the data more closely, with a score that indicated a much tighter fit. Perhaps most importantly, the model was able to correctly identify the strength of the memory effect. The artificial heart was built with a specific memory value of 0.82, and the model estimated it to be 0.83, a very close match. When the researchers ran the test thirty times with different patterns of noise, the model consistently found a value between 0.83 and 0.84, showing that the method is stable and reliable under controlled conditions.

The researchers were careful to explain what their findings do and do not mean. They emphasized that this work is a proof of concept using computer-generated data, not a diagnosis of human patients. The model treats the memory parameter as a way to describe how the shape of the wave behaves, not as a direct measure of a disease or a specific biological process inside the heart muscle. They argue that this parameter is simply a tool to make the model more accurate, much like adjusting the focus on a camera to get a clearer picture. They explicitly ruled out the idea that the heart's internal clock itself needs to be modeled with this complex memory; their tests showed that keeping the clock simple and applying the memory only to the wave shape was the most effective approach.

Looking ahead, the team has laid out a strict plan to see if this method works on real human data. They have specified exactly how they will test the model using public databases of heart recordings, such as the MIT-BIH Arrhythmia Database and the QT Database. They plan to split the data so that the model learns from some patients and is tested on others, ensuring that the results are not just a lucky guess for one specific person. They intend to check if the memory parameter remains consistent across different heartbeats from the same person and if it helps improve the reconstruction of specific parts of the wave, like the recovery phase after the main spike. Until these tests on real human records are completed and published, the researchers caution that the memory parameter should be viewed as a mathematical feature of their model, not as a clinical sign of illness.

This work represents a shift in how scientists might approach the modeling of biological rhythms. Instead of trying to force every part of the system into a single, complex framework, the researchers found success by separating the timing mechanism from the response mechanism. By giving the response a limited, fading memory, they created a model that is both mathematically sound and highly accurate in its predictions. The study demonstrates that even in a field as established as electrocardiography, there is still room to refine how we describe the flow of time and memory in the human body, provided we are willing to test our ideas with rigorous, reproducible methods before claiming they apply to the real world.