Home Journals Articles Updates About
CN
image
收藏切换
Comparative evaluation of Mamdani and Takagi-Sugeno fuzzy systems for predicting peripheral vascular resistance from photoplethysmogram signals using Persian medicine pulsology
收藏切换
PDF
Mohammad Dehghandar1, *, Maryam Nikbakht1, Fatemeh Yousefi Nadarabadi2
Digital Chinese Medicine | 2026, 9(2) : 257 - 264
Less
收藏切换
Digital Chinese Medicine | 2026, 9(2): 257-264
• Original Research •
Comparative evaluation of Mamdani and Takagi-Sugeno fuzzy systems for predicting peripheral vascular resistance from photoplethysmogram signals using Persian medicine pulsology
Full
Mohammad Dehghandar1, *, Maryam Nikbakht1, Fatemeh Yousefi Nadarabadi2
Affiliations
  • 1Department of Mathematics, Payame Noor University, Tehran 3697-19395, Iran
  • 2Applied Mathematics, Payame Noor University, Tehran 3697-19395, Iran
About Author:

Author contributions

Mohammad Dehghandar: conceptualization, methodology, data curation, software, formal analysis, investigation, writing − original draft, supervision, validation, and visualization. Maryam Nikbakht: methodology, software, and data curation. Fatemeh Yousefi Nadarabadi: conceptualization, resources, supervision, project administration, and validation. All authors approved the submission and take responsibility for this manuscript.

Published: 2026-06-25 doi: 10.1016/j.dcmed.2026.05.008
Outline
收藏切换
Objective

To perform a comparative evaluation of Mamdani and Takagi-Sugeno (TS) fuzzy inference systems for predicting peripheral vascular resistance from photoplethysmogram (PPG) signals, incorporating diagnostic parameters from Persian medicine (PM) pulsology.

Methods

Both fuzzy inference systems were implemented in MATLAB R2021b and validated using leave-one-out cross-validation (LOOCV) on clinical PM pulse diagnostic data collected from 35 healthy volunteers. The dataset included pulse frequency scale (1 – 6) and weakness scale (1 – 4) alongside corresponding PPG-derived peripheral vascular resistance indices (range: 0.019 – 0.983). The Mamdani system (with 35 rules) was designed using trapezoidal and triangular membership functions, a singleton fuzzifier, a product inference engine, and a centroid defuzzifier. The TS system (with 6 rules) was configured as a first-order model with Gaussian input membership functions. System performance was quantitatively evaluated using mean absolute error (MAE), root mean square error (RMSE), and the coefficient of determination (R2). Statistical significance of performance differences was assessed using a paired t test.

Results

The Mamdani inference system showed much higher prediction accuracy than the TS system. Comparative analysis revealed a substantial advantage for the Mamdani system (MAE = 0.007 63 ± 0.001 20, RMSE = 0.008 58 ± 0.001 50, R2 = 0.998 40 ± 0.000 80) over the TS system (MAE = 0.015 40 ± 0.002 10, RMSE = 0.022 48 ± 0.002 80, R2 = 0.989 20 ± 0.001 50). A paired t test comparing the absolute errors of the 35 LOOCV folds confirmed statistical significance [t (34) = 5.07, P < 0.001].

Conclusion

The findings suggest the potential suitability of the Mamdani system for certain precision-oriented analytical tasks. Both systems exhibit practical utility for integrative medicine diagnostics and wearable health-monitoring applications for healthy individuals, enabling a modern computational translation of PM pulsology.

Photoplethysmogram  /  Fuzzy inference systems  /  Peripheral vascular resistance  /  Persian medicine  /  Pulse diagnosis  /  Takagi-Sugeno
Mohammad Dehghandar, Maryam Nikbakht, Fatemeh Yousefi Nadarabadi. Comparative evaluation of Mamdani and Takagi-Sugeno fuzzy systems for predicting peripheral vascular resistance from photoplethysmogram signals using Persian medicine pulsology[J]. Digital Chinese Medicine, 2026 , 9 (2) : 257 -264 . DOI: 10.1016/j.dcmed.2026.05.008
Full Text
Less
1　Introduction
Less
Recent advances in computational medicine have highlighted the role of intelligent systems in integrating traditional diagnostic methods with modern technologies. Among these, Persian medicine (PM) pulsology, a centuries-old system based on multidimensional pulse analysis has gained renewed attention for complementing conventional cardiovascular assessments [1-3]. Photoplethysmography (PPG) enables noninvasive hemodynamic monitoring via an optical sensor [4, 5], and the PPG waveform encodes key hemodynamic information, including peripheral vascular resistance [6, 7]. However, the fusion of PM pulsology and PPG faces several key challenges. PM diagnostics are subjective and depend heavily on practitioner expertise [8-11]; automated PPG interpretation remains incomplete [12, 13]; and a unified framework is lacking, as existing studies focus separately on PM classification [1, 11, 14, 15] or PPG feature extraction [16-19]. Fuzzy inference systems can address these challenges through graded membership functions that accommodate subjective pulse descriptors such as “weak” or “strong” [20-22]. Both the Mamdani model, known for its interpretable rule structure, and the Takagi-Sugeno (TS) model, recognized for its adaptive functional consequences, are promising for complex physiological signal analysis [23, 24].
Previous studies have demonstrated the feasibility of Mamdani-type fuzzy systems for estimating PPG features from PM pulse parameters [25-28]. However, these implementations lacked formal stability analysis and systematic performance comparisons between Mamdani and TS architectures. Although the Mamdani versus TS debate has been explored in other domains [29, 30], no dedicated comparison exists for PM-PPG applications.
This study pioneers the application of TS fuzzy inference in PM pulsology beyond the exclusive use of Mamdani models, introduces guaranteed stability modeling absent in prior works, and establishes quantitative correlations between PM pulse parameters (e.g., pulse weakness and frequency) and PPG-derived peripheral vascular resistance. The key innovation is a stable fuzzy framework with bounded outputs. Therefore, this study performs a comparative evaluation of Mamdani and TS fuzzy inference systems for predicting peripheral vascular resistance from PPG signals using PM pulsology.
2　Materials and methods
Less
The methodology of this study, shown in Figure 1, was designed to evaluate the performance of TS and Mamdani fuzzy systems in estimating peripheral vascular resistance from pulse frequency and pulse weakness parameters of PM. The TS system is described and presented in this study, while the Mamdani system was previously used in reference [27].
Several factors motivate the need for fuzzy systems in this context. First, the interpretation of pulse findings by physicians involves subjective experience and analysis, which are prone to errors. Second, the training of PM students also introduces variability in pulse assessment. Moreover, inference based on a large number of pulse rules is inherently complex. Fuzzy systems offer a suitable framework to manage these challenges [25, 28]. Consequently, it is necessary to evaluate the performance quality and error magnitude of different fuzzy systems to select the most appropriate one for a given study. Such evaluation requires a set of actual data, corresponding predicted data, and a performance metric that quantifies how closely the predicted values approximate the actual values.
2.1　Preparing the conditions for designing fuzzy systems
To design a fuzzy system for estimating peripheral vascular resistance using PM pulse frequency and pulse weakness, a peripheral vascular resistance index was first derived from the PPG signal. The preprocessing procedure consisted of the following steps. The area under the PPG curve within a single cardiac cycle was defined as the pulse area. The pulse area was divided into two parts at the dicrotic notch, as shown in Figure 2.
The ratio of the two areas $ {A}_{1} $ and $ {A}_{2} $ shown in Figure 2 is defined as the total peripheral vascular resistance index, given by Equation (1) and referred to as the IPA ratio [16]:
$ \mathrm{IPA}=\frac{{A}_{2}}{{A}_{1}} $
Data were collected at the Ahmadiyeh Persian Medicine Clinic, Tehran University of Medical Sciences, Tehran, Iran, from June 1, 2024 to October 31, 2025. Inclusion criteria for participants were aged between 7 and 69 years, with no known history of cardiovascular disease, and not taking any medication affecting heart rate or vascular tone. All participants (or their legal guardians) provided written informed consent before data collection, in accordance with a protocol approved by the Ethics Committee of Payame Noor University (IR.PNU.REC.1403.685).
The collected data included pulse frequency, pulse weakness, and PPG waveforms, all recorded by a PM specialist. Because the recorded pulse strength was predominantly low (i.e., most pulses were classified as weak), the variable was labeled “pulse weakness” rather than “pulse strength”. Consequently, a decrease in pulse weakness corresponds to an increase in pulse strength, and vice versa. The total peripheral vascular resistance index data were normalized using Equation (2) to reduce computational errors:
$ {x}_{{\mathrm{new}}}=\frac{{x}_{{\mathrm{old}}}-{x}_{{\mathrm{min}}}}{{x}_{{\mathrm{max}}}-{x}_{{\mathrm{min}}}} $
where $ {x}_{{\mathrm{old}}} $ is the original data, $ {x}_{{\mathrm{max}}} $ and $ {x}_{{\mathrm{min}}} $ are the maximum and minimum values in the dataset, respectively, and $ {x}_{{\mathrm{new}}} $ is the normalized data. Normalization reduces scaling-related noise but does not artificially inflate correlations, as the original unnormalized trends remain consistent.
The variables are named as follows: pulse frequency, pulse weakness, and total peripheral vascular resistance index (IPA). These data are given in Table 1.
The steps of designing a stable fuzzy system with two inputs (F, W) and one output (IPA) are as follows.
First, fuzzy sets are defined to cover the input and output spaces. Specifically, fuzzy sets for pulse frequency (F) are defined on the interval [1, 6]. Fuzzy sets for pulse weakness (W) are defined on the interval [1, 4]. Fuzzy sets for the output (peripheral vascular resistance index IPA) are defined on the interval [0, 1]. Next, fuzzy if-then rules are created using input-output data. For each input-output triplet $ \left({F}^{l},{W}^{l};{{\mathrm{IPA}}}^{l}\right) $, the corresponding membership values in the fuzzy sets are determined.
To ensure stability conditions, the following requirements are imposed. First, regarding completeness, a fuzzy set $ {A}_{i} $ is considered complete if Equation (3) holds:
$ {\forall x}_{i}\in \left[\alpha ,\;\beta \right]\Rightarrow {\mu }_{{{A}_{i}}}\left({x}_{i}\right)\neq 0 $
In other words, for every input value within the fuzzy set’s range, the membership function value is non-zero. Second, regarding normality, a fuzzy set $ {A}_{i} $ is normal if Equation (4) holds:
$ {\exists {{x}_{i}}|\mu }_{{{A}_{i}}}\left({x}_{i}\right)=1 $
In other words, there exists at least one point in the fuzzy set where the membership function equals one. Third, regarding consistency, a set of fuzzy if-then rules is consistent if no two rules share the same antecedent (if-part) but have different consequents (then-part).
In this study, “guaranteed stability” refers to bounded-input bounded-output (BIBO) stability. Specifically, for any input values within the defined ranges (F ∈ [1, 6]; W ∈ [1, 4]), the fuzzy system’s output (IPA) will always remain within the predefined output interval [0, 1]. This property is ensured by using complete, normal, and consistent fuzzy sets, as defined in Equations (3) and (4).
2.2　Mamdani fuzzy system
As mentioned in previous study [27], 35 fuzzy rules were formulated based on the defined fuzzy sets and membership functions. The Mamdani fuzzy system was implemented in MATLAB R2021b using these 35 rules, with trapezoidal and triangular membership functions, a singleton fuzzifier, a product inference engine, and a centroid defuzzifier. This system is stable, with input ranges of pulse frequency [1, 6], pulse weakness [1, 4], and an output (IPA) range [0, 1]. These 35 rules were derived directly from expert knowledge provided by a certified PM practitioner, based on established clinical relationships between pulse parameters (frequency and strength) and peripheral vascular resistance (note that pulse weakness is inversely related to pulse strength). The rules were manually formulated rather than automatically extracted from the dataset, thereby ensuring clinical interpretability and face validity.
In this Mamdani fuzzy system, peripheral vascular resistance decreases as both pulse frequency and pulse weakness decrease, and increases as both pulse frequency and pulse weakness increase.
2.3　TS fuzzy system
In a TS fuzzy system, the output of each rule is a first-order polynomial of the input variables. Specifically, consider a TS fuzzy system with r rules, n inputs, and one output. The L-th rule (L = 1, 2, ..., r) is given by Equation (5) [2, 30]:
$\begin{split} {\rm{Rule}}\;{L}: & \;{\rm{if}}\;{x}_{1}\;{\rm{is}}\;A_{1}^{L},\;{\rm{and}}\;{x}_{2}\;{\rm{is}}\;A_{2}^{L},\;{\mathrm{and}}...,{x}_{n}\;{\rm{is}}\;A_{n}^{L},\;{\rm{then}}\\ &{y}^{L}= \alpha _{0}^{L}+\alpha _{1}^{L}{x}_{1}+...+\alpha _{n}^{L}{x}_{n} \end{split}$
where in $ L=\left\{1,2,...,r\right\}, $ $ {\alpha }^{L}=\left\{\alpha _{0}^{L},\alpha _{1}^{L},\alpha _{2}^{L},...,\alpha _{n}^{L}\right\} $ are the consequent parameters. The fuzzy sets for each input are defined as:
$ {A}_{i}=\left\{A_{i}^{1},A_{i}^{2},...,A_{i}^{L}\right\} $
Given an input vector $ x={\left({x}_{1},{x}_{2},...,{x}_{n}\right)}^{T} $, the overall output f (x) is the weighted average of the rule outputs:
$ f(x)=\sum\limits_{L=1}^{r}{y}^{L}{\overline{w}}^{L} $
where $ {w}^{L}=\prod\limits_{i=1}^{n}{\mu }_{{A_{i}^{L}}}\left({x}_{i}\right),\;\;{\overline{w}}^{L}=\dfrac{{w}^{L}}{\displaystyle\sum\limits_{L=1}^{r}{w}^{L}} $.
In this study, the membership functions were chosen to be Gaussian, as expressed in Equation (7) [2]:
$ {\mu }_{{{A}_{i}}}\left({x}_{i}\right)={e}^{-\tfrac{1}{2}{{\left(\tfrac{{x}_{i}-{m}_{i}}{{\sigma }_{i}}\right)}^{2}}} $
where $ {\sigma }_{i} $ and $ {m}_{i} $ are the variance and center (adjustable parameters) of the Gaussian function, respectively.
A TS fuzzy system was designed using MATLAB R2021b based on data from volunteers aged 7 to 69 years. First, fuzzy sets were defined for pulse frequency (F) on the interval [1, 6], pulse weakness (W) on [1, 4], and the IPA output on [0, 1]. Pulse frequency and pulse weakness were used as input variables, and the IPA as the output variable. Using the input-output triplets $ \left({F}^{l},{W}^{l};{{\mathrm{IPA}}}^{l}\right) $ from the volunteers, a total of 6 rules were defined.
The TS system employed a first-order polynomial consequent for each rule, allowing each rule to define a local linear model over a specific region of the input space. Consequently, the model’s expressiveness and adaptability are achieved through the combination of these few local linear functions and the Gaussian antecedent membership functions, rather than through a large number of independent linguistic rules. This characteristic of TS systems typically leads to a more compact rule base compared to the Mamdani architecture, which requires a separate linguistic rule for each distinct input-output mapping [30]. Therefore, while the Mamdani system required 35 rules to linguistically encode expert knowledge, the TS system achieved a comparable mapping of the input-output relationship with only 6 locally linear rules.
The surface diagram illustrating the three-dimensional relationship between the inputs (pulse frequency and pulse weakness) and the output (IPA) of the TS fuzzy system is shown in Figure 3.
2.4　Performance evaluation
Performance evaluation was primarily based on prediction errors. Visual analysis of the results was also performed to support the quantitative assessment.
To evaluate prediction accuracy, the following metrics were calculated: absolute error (AE), mean absolute error (MAE), relative error (RE), mean square error (MSE), root mean square error (RMSE), and coefficient of determination (R2). These metrics are defined in Equations (8) – (13), respectively [29, 31].
$ \mathrm{A}{\mathrm{E}}^{i}=\left| {\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}\right| $
$ \mathrm{MAE}=\frac{1}{N}\sum\limits_{i=1}^{N}\left| {\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}\right| $
$ \mathrm{R}{\mathrm{E}}^{i}=\frac{\left| {\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}\right| }{\left| {\mathrm{IPA}}_{O}^{i}\right| } $
$ \mathrm{MSE}=\frac{1}{N}\sum\limits_{i=1}^{N}{\left({{\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}}\right)}^{2} $
$ \mathrm{RMSE}=\sqrt{\frac{1}{N}\sum\limits_{i=1}^{N}{\left({{\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}}\right)}^{2}} $
$ {R}^{2}=1-\left\{\frac{\displaystyle\sum\limits_{i=1}^{N}{\left({\mathrm{IPA}}_{O}^{i}-{\mathrm{IPA}}_{{\mathrm{FIS}}}^{i}\right)}^{2}}{\displaystyle\sum\limits_{i=1}^{N}{\left({\mathrm{IPA}}_{O}^{i}-\overline{{\mathrm{IPA}}}_{O}^{i}\right)}^{2}}\right\} $
where $ {\mathrm{IPA}}_{O}^{i} $ and $ {\mathrm{IPA}}_{{\mathrm{FIS}}}^{i} $ are the i-th observed and predicted IPA values, respectively, and N is the total number of data points (N = 35).
2.5　Statistical analysis
Statistical analyses were performed using MATLAB R2021b and Microsoft Excel. To robustly estimate prediction accuracy, we calculated the MAE, RMSE, and R2 [defined in Equations (8) – (13)] using leave-one-out cross-validation (LOOCV), which was necessitated by the limited sample size (N = 35). In each LOOCV fold, 34 samples were used for training and the remaining one for testing; this process was repeated 35 times so that each sample served once as the test set. All reported metrics are presented as mean ± SD across the 35 LOOCV folds.
To statistically compare the prediction errors of the Mamdani and TS systems, a paired two-tailed t test was performed on the absolute prediction errors obtained from the 35 LOOCV test folds.
3　Results
Less
The dataset from 35 healthy volunteers included pulse frequency scale (1 − 6) and weakness scale (1 − 4) along with corresponding PPG-derived peripheral vascular resistance indices (range: 0.019 − 0.983).
3.1　AE
The AE results show that both the TS fuzzy system (6 rules) and the Mamdani fuzzy system (35 rules) predict IPA values for 35 healthy volunteers with an error of less than 0.08. Visual analysis of the AEs confirmed that both systems—using pulse frequency and pulse weakness as inputs and IPA as the output—performed satisfactorily within the defined input ranges. The Mamdani fuzzy system achieves a lower absolute error in predicting IPA than the TS fuzzy system. Specifically, the absolute errors of the Mamdani system were lower than those of the TS system for most volunteers, with only a few exceptions (Figure 4).
3.2　RE
The RE results for the two fuzzy systems Mamdani and TS for predicting IPA values in 35 healthy volunteers are shown in Figure 5. As observed with AE, the Mamdani system achieved a substantially lower relative error in predicting IPA than the TS system. This pattern held for most volunteers, with only a few exceptions.
3.3　Regression
Figure 6 shows the regression plots for Mamdani and TS fuzzy systems for 35 healthy volunteers. Visual analysis indicated that, although the TS system yielded better IPA predictions for some volunteers, the Mamdani system was more accurate for the majority of volunteers.
3.4　Simulation results
The simulated results for the two fuzzy inference systems Mamdani and TS for estimating the IPA index from PM pulse parameters can be seen in Figure 7, which illustrates the predicted versus actual IPA values for both systems.
All performance metrics (MAE, RMSE, and R2) reported in Table 2 represent the results obtained across all LOOCV folds.
A paired t test comparing the absolute errors of the 35 LOOCV folds confirmed statistical significance [t (34) = 5.07, P < 0.001].
4　Discussion
Less
4.1　Mamdani’s performance and physiological basis
There are key differences in the performance and usability of the two fuzzy systems, Mamdani and TS. The Mamdani fuzzy system produces outputs by aggregating multiple fuzzy rule outputs using centroid defuzzification [20, 21, 27], whereas each rule in the TS fuzzy system contributes a weighted linear function to the overall output [2, 29, 30]. The results of this study demonstrate that the Mamdani fuzzy system outperformed the TS system across all key metrics. The LOOCV ensured reliable performance estimation despite the limited sample size (N = 35); all metrics are aggregated cross-validation results. This superior performance is evidenced by significantly lower errors. The Mamdani system achieved a MAE of 0.007 63 and a RMSE of 0.008 58, compared to 0.015 40 and 0.022 48 for the TS system, indicating higher predicting precision. Furthermore, the Mamdani system exhibited a near-perfect R2 value of 0.99840, suggesting that its rules align almost perfectly with the observed physiological relationships. Several factors explain this superior performance. The strong physiological correlation between PM pulse parameters (frequency and weakness) and peripheral vascular resistance is well documented in PM literature. The linguistic rule-based architecture of the Mamdani system directly encodes expert diagnostic logic, minimizing model error. Additionally, the centroid defuzzification method employed in the Mamdani system minimizes output variability, a critical feature for medical applications where small errors can significantly impact clinical decisions.
The Mamdani system’s 35 transparent rules, which utilize interpretable membership functions (e.g., triangular and trapezoidal), allow physicians to trace how input variables are mapped to IPA estimates. This aligns perfectly with Persian medicine’s emphasis on interpretable diagnostics. In contrast, the TS system’s crisp linear outputs, despite their computational efficiency, obscure the underlying physiological logic, making it less suitable for clinical validation. The Mamdani system also demonstrated a critical advantage: bounded input-output stability. This feature, ensured by our formal analysis and supported by prior research [27], guarantees that all predictions remain within physiologically plausible ranges. Furthermore, unlike the TS model, which exhibited disturbances during the early design stages, the Mamdani architecture was entirely free of such instabilities [29].
4.2　Stability and interpretability analysis
This study’s contributions are clearly positioned by a direct comparison with key recent works in PPG analysis and computational PM. Our work moves decisively beyond proof-of-concept demonstrations to provide a rigorous, stability-aware architectural comparison. In comparison to foundational PM-PPG studies employing Mamdani-type fuzzy systems, such as those estimating systolic area [25] or peripheral vascular resistance [27], our research introduces two critical advancements. First, we provide the first head-to-head performance evaluation against the TS architecture for this specific physiological prediction task. Second, our implemented Mamdani model achieves superior quantitative accuracy (R2 = 0.998 40, MAE = 0.007 63), exceeding the error thresholds (e.g., MAE < 0.05 [28] or max error 0.01 [27]) reported in these prior single-architecture studies, while formally guaranteeing system stability.
Our findings contribute to the broader debate on fuzzy system selection. In domains like traffic calibration, TS systems have shown superiority for adaptive, real-time control [29]. Conversely, in medical diagnostics, the interpretability of Mamdani systems is often paramount [30]. Our results provide strong, domain-specific evidence. For translating subjective PM diagnostics into quantitative physiological estimates, the Mamdani architecture is not only more interpretable but can also achieve higher predictive accuracy than a first-order TS model.
4.3　Limitations and future study
Several limitations should be acknowledged. The study’s results are based on data from 35 healthy volunteers; therefore, the system’s performance on pathological cases, such as individuals with cardiovascular diseases, requires further validation. Additionally, while the Mamdani system’s defuzzification process carries a higher computational load than the TS system’s weighted averaging, this difference is negligible for offline clinical analysis. Furthermore, the current study was conducted exclusively on healthy volunteers. Therefore, the system’s diagnostic performance in patients with cardiovascular disorders has not yet been established and requires future investigation.
For future research, two primary directions are proposed. First, hybrid systems should be developed that combine the interpretability of the Mamdani framework with the computational efficiency of the TS model, creating architectures suitable for real-time monitoring applications such as wearable PPG devices. Second, the system’s performance must be validated on pathological cases, particularly patients with vascular disorders, to thoroughly assess its clinical generalizability and diagnostic robustness beyond healthy populations.
5　Conclusion
Less
This study successfully developed a stable and interpretable fuzzy system for predicting the peripheral vascular resistance index (IPA) from PM pulse parameters. The Mamdani system emerged as the superior framework, supported by its exceptional predictive accuracy: a near-perfect R2 value of 0.99840 and low prediction errors (MAE: 0.00763). This validates its precise alignment with established PM pulsology principles. Furthermore, its transparency allows physicians to examine all 35 rules to understand how specific pulse features (e.g., frequency and weakness) map to the IPA output, thereby fostering trust in artificial intelligence (AI)-assisted diagnostics. The system’s inherent safety, guaranteed by bounded output stability, ensures that all predictions remain within physiologically plausible ranges. The Mamdani framework offers high accuracy and interpretability for estimating peripheral vascular resistance from PM pulse parameters in healthy individuals.
References
Less
1
ALIZADEH M, KESHAVARZ M, EBADIANI M, et al. Complexity and rationality of Avicenna’s pulsology: a step towards understanding the past for today’s applications. International Journal of Cardiology, 2012, 157(3): 434–435.
2
DEHGHANDAR M, REZVANI S. Classification of COVID-19 individuals using adaptive neuro-fuzzy inference system. Journal of Medical Signals & Sensors, 2022, 12(4): 334–340.
3
DEHGHANDAR M. Investigating the effect of pulse traditional medicine on the factors of blood pressure by fuzzy calculations. Tehran: Payame Noor University, 2016.
4
PADILLA JM, BERJANO EJ, SAIZ J, et al. Assessment of relationships between blood pressure, pulse wave velocity, and digital volume pulse. Computers in Cardiology, 2006, 33: 893–896.
5
ASADA HH, SHALTIS P, REISNER A, et al. Mobile monitoring with wearable photoplethysmographic biosensors. IEEE Engineering in Medicine and Biology Magazine, 2003, 22(3): 28–40.
6
PARK J, SEOK HS, KIM SS, et al. Photoplethysmogram analysis and applications: an integrative review. Frontiers in Physiology, 2022, 12: 808451.
7
RAMZAN M, HAMID M, ALI ALHUSSAN A, et al. Accurate prediction of anxiety levels in Asian countries using a fuzzy expert system. Healthcare, 2023, 11(11): 1594.
8
CHASHTI M. Exir-e-Azam (Great Elixir). Tehran: Research Institute for Islamic and Complementary Medicine, 2008.
9
ALIZADEH VAGHASLOO M. Explaining the ten parameters of pulse diagnosis in traditional Iranian medicine. Tehran: Tehran University of Medical Sciences, 2013.
10
DEHGHANDAR M, KHALOOZADEH H, SOLTANIAN F, et al. Modeling of the pulse parameters in Iranian traditional medicine and estimation of blood pressure by an artificial neural network. Journal of Industrial Engineering Research, 2016, 2(1): 17–22.
11
NASERI M, REZAI ZADEH H, CHOOPANI R. General overview of traditional medicine. Tehran: Nashre Shahr, 2010: 14-27.
12
LIU SH, WANG JJ, CHEN WX, et al. Classification of photoplethysmographic signal quality with fuzzy neural network for improvement of stroke volume measurement. Applied Sciences, 2020, 10(4): 1476.
13
MOUSAVI SS, FIROUZMAND M, CHARMI M, et al. Blood pressure estimation from appropriate and inappropriate PPG signals using a whole-based method. Biomedical Signal Processing and Control, 2019, 47: 196–206.
14
NAFISI VR, GHODS R, SHOJAEDINI SV. A novel pulse-taking device for Persian medicine based on convolutional neural networks. Journal of Medical Signals & Sensors, 2022, 12(4): 285–293.
15
NAFISI VR, GHODS R. A telecare system for use in traditional Persian medicine. The Open Biomedical Engineering Journal, 2021, 15(1): 105–114.
16
ELGENDI M. On the analysis of fingertip photoplethysmogram signals. Current Cardiology Reviews, 2012, 8(1): 14–25.
17
LIU SH, LI RX, WANG JJ, et al. Classification of photoplethysmographic signal quality with deep convolution neural networks for accurate measurement of cardiac stroke volume. Applied Sciences, 2020, 10(13): 4612.
18
KIM SC, CHO SH. Blood pressure estimation algorithm based on photoplethysmography pulse analyses. Applied Sciences, 2020, 10(12): 4068.
19
FINNEGAN E, DAVIDSON S, HARFORD M, et al. Features from the photoplethysmogram and the electrocardiogram for estimating changes in blood pressure. Scientific Reports, 2023, 13: 986.
20
MENDEL JM. Explainable Uncertain Rule-Based Fuzzy Systems. Cham: Springer International Publishing, 2024.
21
DEHGHANDAR M. Diagnosis of COVID-19 disease by fuzzy expert system designed based on input-output. Journal of Control, 2021, 14(5): 71–78.
22
VAGHASLOO MA, NAGHIZADEH A, KESHAVARZ M. The concept of pulse. Traditional and Integrative Medicine, 2017, 2(2): 54–60.
23
KUSUMADEWI S, ROSITA L, WAHYUNI EG, et al. Implementation of decision tree and Mamdani fuzzy inference system for Erythropoietin resistance prediction. Biomedical Signal Processing and Control, 2025, 104: 107496.
24
DUBEY P, KUMAR S, BEHERA SK, et al. A Takagi-Sugeno fuzzy controller for minimizing cancer cells with application to androgen deprivation therapy. Healthcare Analytics, 2023, 4: 100277.
25
DEHGHANDAR M, ALIZADEH VAGHASLOO M, MIRHOSSEINI-ALIZAMINI S, et al. Fuzzy controller design for estimation of systolic area in PPG signal by Persian medicine. Digital Diagnostics, 2025, 6(3): 414-426.
26
DEHGHANDAR M, ALIZADEH VAGHASLOO M, MORADI B. Estimation of men’s brachial blood pressure gradient using fuzzy system by pulse parameters in Persian medicine. Proceedings of the 4th National Seminar on Control and Optimization, 2022.
27
DEHGHANDAR M, NIKBAKHT M, YOUSEFI NADARABADI F. Designing a fuzzy stable system to estimate peripheral resistance using frequency and pulse strength parameters of Persian medicine. Journal of Health and Biomedical Informatics, 2025, 11(4): 291–305.
28
DEHGHANDAR M, MIRHOSSEINI-ALIZAMINI M, ALIZADEH VAGHASLOO M, et al. Design and implementation of a fuzzy intelligent system to estimate the photoplethysmogram systolic features by Persian medicine pulsology. Health Management and Information Science, 2024, 10(4): 226–234.
29
POP MD, PESCARU D, MICEA MV. Mamdani vs. Takagi-Sugeno fuzzy inference systems in the calibration of continuous-time car-following models. Sensors, 2023, 23(21): 8791.
30
OUIFAK H, IDRI A. On the performance and interpretability of Mamdani and Takagi-Sugeno-Kang based neuro-fuzzy systems for medical diagnosis. Scientific African, 2023, 20: e01610.
31
XIAO XH, WANG ZP, ZHANG HP, et al. A novel method of bridge deflection prediction using probabilistic deep learning and measured data. Sensors, 2024, 24(21): 6863.
Appendix
Less
Year 2026 volume 9 Issue 2
PDF
110
59
Cite this Article
BibTeX
Article Info
doi: 10.1016/j.dcmed.2026.05.008
  • Receive Date：2025-10-31
  • Online Date：2026-08-20
  • Published：2026-06-25
Article Data
Affiliations
History
  • Received：2025-10-31
  • Accepted：2026-04-16
Affiliations
    1Department of Mathematics, Payame Noor University, Tehran 3697-19395, Iran
    2Applied Mathematics, Payame Noor University, Tehran 3697-19395, Iran

Corresponding:

References
Share
https://castjournals.cast.org.cn/joweb/dcm/EN/10.1016/j.dcmed.2026.05.008
Share to
QR

Scan QR to access full text

Cite this article
BibTeX
Citations
表12种不同金属材料的力学参数
科
Family		 属数
Number of
genus		 种数
Number of
species		 占总种数比例
Percentage of
total species (%)		 属
Genus		 种数
Number of
species		 占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏
  • BibTeX
  • EndNote
  • RefWorks
  • TxT
Links
Articles
Latest Articles
Most Read
Collections
Updates
Events
News
Multimedia
About
About Us
Contact
No. 86 Xueyuan South Road, Haidian District, Beijing
100081
010-62199257
qkjq@cast.org.cn

Copyright © 2025 China Association for Science and Technology. All rights reserved. For all open access content, the relevant licensing terms apply.
Sponsored by the Office of the Leading Group for Cybersecurity and Informatization of CAST, and supported by Science and Technology Review Publishing House