Predicting diameter and tensile strength of electrospun fibers for biomedicine: A comparison of Box-Behnken design, traditional machine learning and deep learning
| Parent link: | Computers in Biology and Medicine.— .— Amsterdam: Elsevier Science Publishing Company Inc. Vol. 196.— 2025.— Article number 110923, 15 p. |
|---|---|
| Other Authors: | , , , |
| Summary: | Title screen Electrospun polycaprolactone scaffolds are used in biomedical tissue engineering due to their biodegradability and mechanical tunability. This study systematically compares the three design approaches Box-Behnken design, non-neural network algorithms and artificial neural networks to predict fiber diameter and tensile strength based on the electrospinning parameters: polymer concentration, applied voltage, and needle size. This represents a novel analysis of electrospun scaffolds with significantly distinct morphologies, which has not yet been thoroughly investigated before. These electrospinning parameters were varied to fabricate 18 electrospun scaffold samples following the Box-Behnken design with 15 training and 3 test samples to obtain experimental fiber diameter and tensile strength values. The fiber diameter and tensile strength were predicted through 180 non-neural network algorithms, including linear regression, support vector regression, k-nearest neighbors regression, random forest regression, gradient boosting regression and extreme gradient boosting regression and 480 artificial neural network models. The results revealed that the Box-Behnken design models were overfitting and could not predict fiber diameter and tensile strength data. However, the two best non-neural network algorithm models demonstrated precise predictions for the fiber diameter, but encountered problems with the tensile strength. The two best artificial neural network models were able to accurately predict fiber diameters and provide reliable predictions for tensile strength compared to the Box-Behnken design and non-neural network algorithm predictions. The robustness of artificial neural network models was validated through the successful prediction of literature data on fiber diameter and tensile strength of electrospun polymer scaffolds with distinct morphology, confirming generalizability beyond experimental conditions. This work establishes that properly configured artificial neural network models can extract meaningful patterns from small experimental datasets (n ≥ 18) with complex dependencies, thereby reducing the need for extensive experimentation. The final results emphasize the success of artificial neural network models for analyzing and predicting physical parameters even for scaffolds with significantly different morphologies Текстовый файл AM_Agreement |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://doi.org/10.1016/j.compbiomed.2025.110923 |
| Format: | Electronic Book Chapter |
| KOHA link: | https://koha.lib.tpu.ru/cgi-bin/koha/opac-detail.pl?biblionumber=681741 |
MARC
| LEADER | 00000naa0a2200000 4500 | ||
|---|---|---|---|
| 001 | 681741 | ||
| 005 | 20250923114016.0 | ||
| 090 | |a 681741 | ||
| 100 | |a 20250923d2025 k||y0rusy50 ba | ||
| 101 | 0 | |a eng | |
| 102 | |a NL | ||
| 135 | |a drcn ---uucaa | ||
| 181 | 0 | |a i |b e | |
| 182 | 0 | |a b | |
| 183 | 0 | |a cr |2 RDAcarrier | |
| 200 | 1 | |a Predicting diameter and tensile strength of electrospun fibers for biomedicine: A comparison of Box-Behnken design, traditional machine learning and deep learning |f Arsalan D. Badaraev, Sven Rutkowski, Shadfar Davoodi, Sergei I. Tverdokhlebov | |
| 203 | |a Текст |b визуальный |c электронный | ||
| 283 | |a online_resource |2 RDAcarrier | ||
| 300 | |a Title screen | ||
| 320 | |a References: 50 tit | ||
| 330 | |a Electrospun polycaprolactone scaffolds are used in biomedical tissue engineering due to their biodegradability and mechanical tunability. This study systematically compares the three design approaches Box-Behnken design, non-neural network algorithms and artificial neural networks to predict fiber diameter and tensile strength based on the electrospinning parameters: polymer concentration, applied voltage, and needle size. This represents a novel analysis of electrospun scaffolds with significantly distinct morphologies, which has not yet been thoroughly investigated before. These electrospinning parameters were varied to fabricate 18 electrospun scaffold samples following the Box-Behnken design with 15 training and 3 test samples to obtain experimental fiber diameter and tensile strength values. The fiber diameter and tensile strength were predicted through 180 non-neural network algorithms, including linear regression, support vector regression, k-nearest neighbors regression, random forest regression, gradient boosting regression and extreme gradient boosting regression and 480 artificial neural network models. The results revealed that the Box-Behnken design models were overfitting and could not predict fiber diameter and tensile strength data. However, the two best non-neural network algorithm models demonstrated precise predictions for the fiber diameter, but encountered problems with the tensile strength. The two best artificial neural network models were able to accurately predict fiber diameters and provide reliable predictions for tensile strength compared to the Box-Behnken design and non-neural network algorithm predictions. The robustness of artificial neural network models was validated through the successful prediction of literature data on fiber diameter and tensile strength of electrospun polymer scaffolds with distinct morphology, confirming generalizability beyond experimental conditions. This work establishes that properly configured artificial neural network models can extract meaningful patterns from small experimental datasets (n ≥ 18) with complex dependencies, thereby reducing the need for extensive experimentation. The final results emphasize the success of artificial neural network models for analyzing and predicting physical parameters even for scaffolds with significantly different morphologies | ||
| 336 | |a Текстовый файл | ||
| 371 | 0 | |a AM_Agreement | |
| 461 | 1 | |t Computers in Biology and Medicine |c Amsterdam |n Elsevier Science Publishing Company Inc. | |
| 463 | 1 | |t Vol. 196 |v Article number 110923, 15 p. |d 2025 | |
| 610 | 1 | |a электронный ресурс | |
| 610 | 1 | |a труды учёных ТПУ | |
| 610 | 1 | |a Electrospun scaffolds | |
| 610 | 1 | |a Box-Behnken design | |
| 610 | 1 | |a Linear regression | |
| 610 | 1 | |a Ensemble methods | |
| 610 | 1 | |a Artificial neural network | |
| 610 | 1 | |a Non-neural network algorithm | |
| 701 | 1 | |a Badaraev |b A. D. |c Physicist |c Engineer of Tomsk Polytechnic University |f 1995- |g Arsalan Dorzhievich |9 22441 | |
| 701 | 1 | |a Rutkowski |b S. |c chemist |c Research Engineer, Tomsk Polytechnic University, Ph.D |f 1981- |g Sven |9 22409 | |
| 701 | 1 | |a Davoodi |b Sh. |c specialist in the field of petroleum engineering |c Research Engineer of Tomsk Polytechnic University |f 1990- |g Shadfar |9 22200 | |
| 701 | 1 | |a Tverdokhlebov |b S. I. |c physicist |c Associate Professor of Tomsk Polytechnic University, Candidate of physical and mathematical science |f 1961- |g Sergei Ivanovich |9 15101 | |
| 801 | 0 | |a RU |b 63413507 |c 20250923 | |
| 850 | |a 63413507 | ||
| 856 | 4 | |u https://doi.org/10.1016/j.compbiomed.2025.110923 |z https://doi.org/10.1016/j.compbiomed.2025.110923 | |
| 942 | |c CF | ||