A Review for Machine Learning Applications in Characterizing Biomaterials and Biological Materials Properties

Introduction

Biological materials are natural biocompatible materials present inside living organisms, while biomaterials are typically synthetic and engineered to interact with biological systems. Characterizing mechanical properties of biological materials is crucial for disease diagnosis because the alteration of mechanical properties directly relates to disease progression. For instance, calcified human leaflet becomes severe stiff, leading to aortic stenosis and the cell disease are related to increasing stiffness and adhesiveness [1- 2]. Furthermore, the mechanical properties of biomaterials affect their interaction with biological systems [3]. The mechanical properties generally imply elasticity, stiffness and viscosity, quantitatively extracted from experimental measurements. The length scale of biological material and biomaterials can range from nanometer to millimeters, beyond the feasibility of traditional tension and compression tests. Therefore, multiple advanced probing techniques have been developed and accomplished great success in the past. For instance, Atomic Force Microscopy (AFM) and Optical Tweezers (OT) are utilized in nano- and micrometerscale detection [4-5]. AFM technique is usually used to measure the cell and tissue viscoelastic properties by correlating the material deformation and the force applied to the indenters. OT utilizes the laser to generate the repulsive force acting on the cell materials and obtain specific mechanical properties. Although those techniques can reach a substantially small scale, they are destructive testing and hardly maintain the physiological environment for cells and tissue. Moreover, the measurement can only perform at singlecell and tissue levels, which is highly time-consuming for cell and tissue screening. Alternatively, Confocal Microscopy, Coherence Tomography (OCT) and Optical Coherence Elastography (OCE) can generate measurement images in micrometers scales based on the interferometry principle [6-8]. Their applications are limited in the regime between cells and organs; those techniques generally suffer from speckle decorrelation noise, especially for dense tissue materials [9]. Moreover, the living human tissue is limited by availability. Magnetic Resonance Imaging (MRI), Computed Tomography (CT) and ultrasound can perform millimeter-scale measurements to reconstruct the 3D geometric image for further diagnosis and they are non-invasive techniques [10-12]. However, those techniques typically record the material deformation in the vicinity of body fluid and perform a reverse calculation for material properties, such that the knowledge of relative boundary conditions is required. The non- simultaneously measurement of body fluid will also introduce more system uncertainty toward material characterization. Moreover, the rapid development of biomaterials requires the realization of mechanical properties and the understanding of their biocompatibility and their performance within biological systems. For instance, the toxicity of metallic biomaterial, chemical and thermal properties of ceramic biomaterials is within this field’s scope [13]. All these requirements bring extra challenges to current experimental measurement.

Alternatively, Machine learning algorithms can overcome those limitations within current experimental techniques. They are more cost-friendly and can achieve a fast analysis of a large amount of data. The data source can either be obtained by experimental measurement or collected from the online database or previous reports. ML has progressively fascinated many researchers over the past two decades, from the application in computer science at the very beginning to broad disciplines such as engineering, biological and medical researches. There are broadly two major categories depending on learning style: supervised and unsupervised learning; while supervised learning requires user- defined labeling for input data, unsupervised learning demands the computer to discover hidden patterns from un-labeled data [14,15]. An ML procedure generally includes input data preparation, model creation, training and predictions. The implementation of the ML model becomes increasingly convenient nowadays with the help of various platforms such as TensorFlow and Sklearn. The most representative ML models include but are not limited to supportvector machines (SVM), Artificial neural networks/Convolutional networks (ANN/CNN), Gaussian processes (GPs), and Genetic algorithms (GA) and Regression analysis, etc. [16-19]. This paper briefly reviews the current state-of-the-art ML algorithms for the applications in biological materials and biomaterials, including multi- physics and multi-scales, natural or synthetic materials.

Machine Learning Application for Biological Materials

Biological materials typically refer to natural materials that are viable and capable of self-repairing. Human cells, tissue and organ are the most common biological materials. The ML approaches at the cell level focus on the characterization and classification of cell membrane properties under various pathophysiological conditions. Several studies have reported the exciting application of ML in Red Blood Cells (RBC) classification [20-24]. Several experimental techniques are utilized for extracting the cell features that provides input data set for ML purposes. Moreover, various cell types are studied by ML classification, such as circulating tumor cells, lymphoid cells, blast and stem cells, etc. [25-30]. The state-ofthe- art ML algorithms, such as Regression analysis, SVM and ANN/ CNN, have been employed and discussed in the references. One of the most significant contributions in their work is utilizing the ML method to elucidate the relation between cell morphology and cell disorders, facilitating an efficient screening for cell diseases. Furthermore, ML approaches are applied to estimate the specific elastic properties of the material at the tissue scale, such as the vascular and collagen tissues [31,32]. The random forest regression algorithm and ANN are also utilized to invert the Monte Carlo simulations for the extraction of tissue optical properties [33-35] and improve the awareness of tissue damage mechanisms.

At the organ level, CNN using radio frequency data is adopted to understand the liver and breast deformation [36]. The authors considered a mixture of simulation and experimental results as model inputs. K-Means analysis is also applied to predict the real-time breast tissue deformation based on finite element (FE) simulations [37]; a similar study uses integrated ML approach and FE simulations towards breast deformation in [38]. Their works demonstrate the potential of simulation-based ML approach. The human aortic valve (AV) is the most crucial component inside human body; many researchers have harnessed the power of ML techniques for a robust and efficient AV functional characterization. Specifically, ANN has been utilized to derive the specific constitutive model parameters of the aortic wall [39,40]. The mechanical response of the aorta has also been investigated by CNN and regression analysis [41-43]. For Bioprosthetic valves, ML applications are mainly focused on valve geometric reconstruction and critical geometric decision [44-47]. Moreover, research utilizes the autoencoder-based ML method to optimize the design of transcatheter aortic valves and explores valve deformation behavior [48,49]. The main objective of ML applications in AV is to correlate the AV critical geometric information to its functionality and bypass the complex and time-consuming computational analysis and clinical experiments. This type of application substantially facilitates a better understanding of AV function and can assist the physician and clinician in the therapy planning.

Machine Learning Application for Bio-materials

Biomaterials play a crucial role in medical applications, such as drug delivery, tissue engineering and medical diagnostics. Characterization of biomaterial properties becomes significant as their interactions with biological systems are highly dependent on their mechanical properties. Though biomaterials have a wide variety, we mainly discuss the ML applications in three popular biomaterials: polymers, metal and ceramic. One of the most outstanding work in ML application of polymer characterization is the Polymer Genome Project [50]. This informatics platform takes polymer key features as input and predicts the corresponding polymer properties based on similarity in the database using GPs, thus waives the necessity for self-design machine learning networks. Many applications for specific polymer material characterization can also be found in the literature such as dielectric polymeric materials characterization using support vector machine [51]; bacterial cytoskeleton polymetric component using decision tree for classification and regression [52]; Quantitative Structure Activity/Property relationships (QSAR/QSPR) integrated with ANN for the surface properties of biodegradable polymers [53]. Another interesting work is utilizing self- organizing maps (SOM) to classify polymer film types since the biomaterial surface properties control the biomaterial interaction with biological systems [54].

For metallic biomaterials, an attractive application of ML in recent years is in medical implant using the high-entropy alloys (HEAs) [55]; several machine learning methods are employed to classify phase status of HEAs and characterize the materials to targeted properties [56-59]. Besides, machine learning algorithms are emerging as a promising tool for evaluating the cytotoxicity of biomaterial nanoparticles. A unified quantitative structure toxicity relationships (QSTR) perturbation model based on ANN has been proposed to assess general cytotoxicity of the nanoparticles [60]; The Decision tree has been utilized to classify cytotoxicity based on cell viability [61]; Similar work using supervised learning method can also be found in [62,63]. Although those machine learning applications are not a direct characterization of metallic biomaterial properties, they explore the feasibility of the ML method to assess the biocompatibility of biomaterials. Another heavily used material for the implant is ceramics due to their light weight and favorable mechanical properties. Nowadays, researchers focus more on ceramic composites whose components are complex such that the traditional experimental test cannot fully depict their mechanical properties. Instead, machine learning methods can facilitate the development of novel ceramic materials and predict their specific material properties. For instance, linear/nonlinear regression machine learning algorithms and CNN are employed to design bioceramics and bioglass [64,65]. Upon reviewing current progress in the machine learning applications, it is evident that machine learning approaches have a broad range of applications in many aspects of biomaterial development, properties characterization, and material performance assessment.

Summary and Future Expectation

A better understanding of biological materials can also facilitate the development of biomaterials. Meanwhile, ML can assist in the discovery of novel biomaterials. However, the fidelity of a machine learning algorithm is highly dependent on the model architecture, model hyperparameters tuning and input data preparation. The researchers must also carefully validate the integrity of the input parameter metrics to avoid unnecessary system uncertainties. Moreover, machine learning relies on conventional experimental and computational methods for input data collection because it is difficult to establish a comprehensive database for a wide variety of biological materials and biomaterials. However, significant progress has been made in recent years [13]. We expect broader machine learning applications in bio-related material science with the rapid evolution of more sophisticated machine learning networks and more data source availability.

Acknowledgement

The authors are grateful for the support from the National Science Foundation under Grant Number CBET-1706295.

Declaration of Competing Interest

The authors declare no conflicts of interest.

References

• Carabello BA, Paulus WJ (2009) Aortic stenosis. The lancet 373(9667): 956-966.
• Li X, Dao M, Lykotrafitis G, Karniadakis GE (2017) Biomechanics and biorheology of red blood cells in sickle cell anemia. Journal of biomechanics 50: 34-41.
• Lim JY, Donahue HJ (2004) Biomaterial characteristics important to skeletal tissue engineering. Journal of Musculoskeletal and Neuronal Interactions 4(4): 396.
• Akhtar R, Sherratt MJ, Cruickshank JK, Derby B (2011) Characterizing the elastic properties of tissues. Materials Today 14(3): 96-105.
• Moffitt JR, Chemla YR, Smith SB, Bustamante C (2008) Recent advances in optical tweezers. Annu Rev Biochem 77: 205-228.
• Scarcelli G, Yun SH (2008) Confocal Brillouin microscopy for three-dimensional mechanical imaging. Nature photonics 2(1): 39-43.
• Zhu J, Miao Y, Qi L, Qu Y, He Y, et al. (2017) Longitudinal shear wave imaging for elasticity mapping using optical coherence elastography. Applied physics letters 110(20): 201101.
• Kut C, Chaichana KL, Xi J, Raza SM, Ye X, et al. (2015) Detection of human brain cancer infiltration ex vivo and in vivo using quantitative optical coherence tomography. Science translational medicine 7(292): 292ra100-292ra100.
• Mowbray M, Savage T, Wu C, Song Z, Cho BA, et al. (2021) Machine learning for biochemical engineering: A review. Biochemical Engineering Journal.
• Braun J, Guo J, Lützkendorf R, Stadler J, Papazoglou S, et al. (2014) High-resolution mechanical imaging of the human brain by three-dimensional multifrequency magnetic resonance elastography at 7T. Neuroimage 90: 308-314.
• Righetti R, Ophir J, Ktonas P (2002) Axial resolution in elastography. Ultrasound in medicine & biology 28(1): 101-113.
• Buzug TM (2011) Computed tomography. Springer handbook of medical technology, Springer, Berlin, Heidelberg, pp. 311-342.
• Kerner J, Dogan A, von Recum H (2021) Machine Learning and Big Data Provide Crucial Insight for Future Biomaterials Discovery and Research. Acta Biomaterialia.
• Hastie T, Tibshirani R, Friedman J (2009) Overview of supervised learning. The elements of statistical learning, Springer, New York, USA, p. 9-41.
• Hastie T, Tibshirani R, Friedman J (2009) Unsupervised learning. The elements of statistical learning, Springer, New York, USA, pp. 485-585.
• Suthaharan S (2016) Support vector machine. Machine learning models and algorithms for big data classification, Springer, Boston, MA, USA, pp. 207-235.
• Wang SC (2003) Artificial neural network. Interdisciplinary computing in java programming, Springer, Boston, MA, USA, pp. 81-100.
• Rasmussen CE (2003) Gaussian processes in machine learning. Summer school on machine learning, Springer, Berlin, Heidelberg, Germany, p. 63-71.
• De Jong K (1990) Genetic-algorithm-based learning. Machine learning, Morgan Kaufmann, pp. 611-638.
• Kim G, Jo Y, Cho H, Min HS, Park Y (2019) Learning-based screening of hematologic disorders using quantitative phase imaging of individual red blood cells. Biosensors and Bioelectronics 123: 69-76.
• Shirazi, S. H., Umar, A. I., Haq, N., Naz, S., Razzak, M. I., & Zaib, A. (2018). Extreme learning machine based microscopic red blood cells classification. Cluster Computing, 21(1), 691-701.
• Kihm A, Kaestner L, Wagner C, Quint S (2018) Classification of red blood cell shapes in flow using outlier tolerant machine learning. PLoS computational biology 14(6): e1006278.
• Go T, Kim JH, Byeon H, Lee SJ (2018) Machine learning‐based in‐line holographic sensing of unstained malaria‐infected red blood cells. Journal of biophotonics 11(9): e201800101.
• Singh V, Srivastava V, Mehta DS (2020) Machine learning-based screening of red blood cells using quantitative phase imaging with micro-spectrocolorimetry. Optics & Laser Technology 124: 105980.
• Rodellar J, Alférez S, Acevedo A, Molina A, Merino A (2018) Image processing and machine learning in the morphological analysis of blood cells. International journal of laboratory hematology 40(Suppl 1): 46-53.
• Chen D, Sarkar S, Candia J, Florczyk SJ, Bodhak S, et al. (2016) Machine learning based methodology to identify cell shape phenotypes associated with microenvironmental cues. Biomaterials 104: 104-118.
• Iyer A, Gupta K, Sharma S, Hari K, Lee YF, et al. (2020) Integrative analysis and machine learning based characterization of single circulating tumor cells. Journal of clinical medicine 9(4): 1206.
• Nissim N, Dudaie M, Barnea I, Shaked NT (2020) Real‐time stain‐free classification of cancer cells and blood cells using interferometric phase microscopy and machine learning. Cytometry Part A 99(5): 511-523.
• Nassar M, Doan M, Filby A, Wolkenhauer O, Fogg DK, et al. (2019) Label‐free identification of white blood cells using machine learning. Cytometry Part A 95(8): 836-842.
• Darling EM, Guilak F (2008) A neural network model for cell classification based on single-cell biomechanical properties. Tissue Engineering Part A 14(9): 1507-1515.
• Liang L, Liu M, Sun W (2017) A deep learning approach to estimate chemically-treated collagenous tissue nonlinear anisotropic stress-strain responses from microscopy images. Acta biomaterialia 63: 227-235.
• Cilla M, Pérez‐Rey I, Martínez MA, Peña E, Martínez J (2018) On the use of machine learning techniques for the mechanical characterization of soft biological tissues. International journal for numerical methods in biomedical engineering 34(10): e3121.
• Hokr BH, Bixler JN (2021) Machine learning estimation of tissue optical properties. Scientific Reports 11(1): 1-7.
• Doronin A, Meglinski I (2019) The application of a unified Monte Carlo model in the training of artificial neural networks for the purpose of real-time in-vivo sensing of tissue optical properties. Micro-and Nanotechnology Sensors, Systems, and Applications XI, International Society for Optics and Photonics, pp. 109820N.
• Panigrahi S, Gioux S (2018) Machine learning approach for rapid and accurate estimation of optical properties using spatial frequency domain imaging. Journal of biomedical optics 24(7): 071606.
• Wu S, Gao Z, Liu Z, Luo J, Zhang H, et al. (2018) Direct reconstruction of ultrasound elastography using an end-to-end deep neural network. In International conference on medical image computing and computer-assisted intervention, Springer, Cham, pp. 374-382.
• Martínez-Martínez F, Rupérez-Moreno MJ, Martínez-Sober M, Solves-Llorens JA, Lorente D, et al. (2017) A finite element-based machine learning approach for modeling the mechanical behavior of the breast tissues under compression in real-time. Computers in biology and medicine 90: 116-124.
• Martin-Guerrero JD, Ruperez-Moreno MJ, Martinez-Martinez F, Lorente-Garrido D, Serrano-Lopez AJ, et al. (2016) Machine learning for modeling the biomechanical behavior of human soft tissue. In 2016 IEEE 16th International Conference on Data Mining Workshops (ICDMW), IEEE, pp. 247-253.
• Liu M, Liang L, Sun W (2018) Estimation of in vivo constitutive parameters of the aortic wall: a machine learning approach. bioRxiv, 366963.
• Liu M, Liang L, Sun W (2019) Estimation of in vivo constitutive parameters of the aortic wall using a machine learning approach. Computer methods in applied mechanics and engineering 347: 201-217.
• Liang L, Liu M, Sun W (2017) A deep learning approach to estimate chemically-treated collagenous tissue nonlinear anisotropic stress-strain responses from microscopy images. Acta biomaterialia 63: 227-235.
• Liang L, Liu M, Martin C, Sun W (2018) A deep learning approach to estimate stress distribution: a fast and accurate surrogate of finite-element analysis. Journal of The Royal Society Interface 15(138): 20170844.
• Aranda A, Valencia A (2018) Study on cerebral aneurysms: Rupture risk prediction using geometrical parameters and wall shear stress with cfd and machine learning tools. Machine Learning and Applications: An International Journal (MLAIJ).
• Liang L, Kong F, Martin C, Pham T, Wang Q, et al. (2017) Machine learning–based 3‐D geometry reconstruction and modeling of aortic valve deformation using 3‐D computed tomography images. International journal for numerical methods in biomedical engineering 33(5): e2827.
• Hagenah J, Werrmann E, Scharfschwerdt M, Ernst F, Metzner C (2016) Prediction of individual prosthesis size for valve-sparing aortic root reconstruction based on geometric features. In 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), IEEE, pp. 3273-3276.
• Liang L, Martin C, Wang Q, Sun W, Duncan J (2016) Estimation of aortic valve leaflets from 3D CT images using local shape dictionaries and linear coding. In Medical Imaging 2016: Image Processing, International Society for Optics and Photonics, p. 978432.
• Hagenah J, Scharfschwerdt M, Schweikard A, Metzner C (2017) Combining deformation modeling and machine learning for personalized prosthesis size prediction in valve-sparing aortic root reconstruction. In International Conference on Functional Imaging and Modeling of the Heart, Springer, Cham, pp. 461-470.
• Liang L, Sun B (2019) A Proof of Concept Study of Using Machine-Learning in Artificial Aortic Valve Design: From Leaflet Design to Stress Analysis. Bioengineering 6(4): 104.
• Balu A, Nallagonda S, Xu F, Krishnamurthy A, Hsu MC, et al. (2019) A deep learning framework for design and analysis of surgical bioprosthetic heart valves. Scientific reports 9(1): 1-12.
• Kim C, Chandrasekaran A, Huan TD, Das D, Ramprasad R (2018) Polymer genome: a data-powered polymer informatics platform for property predictions. The Journal of Physical Chemistry C 122(31): 17575-17585.
• Wu K, Sukumar N, Lanzillo NA, Wang C, Breneman C, et al. (2016) Prediction of polymer properties using infinite chain descriptors (ICD) and machine learning: Toward optimized dielectric polymeric materials. Journal of Polymer Science Part B: Polymer Physics 54(20): 2082-2091.
• Asgharzadeh P, Birkhold AI, Trivedi Z, Özdemir B, Reski R, et al. (2020) A NanoFE simulation-based surrogate machine learning model to predict mechanical functionality of protein networks from live confocal imaging. Computational and structural biotechnology journal 18: 2774-2788.
• Smith JR, Kholodovych V, Knight D, Welsh WJ, Kohn J (2005) QSAR models for the analysis of bioresponse data from combinatorial libraries of biomaterials. QSAR & Combinatorial Science 24(1): 99-113.
• Madiona RM, Winkler DA, Muir BW, Pigram PJ (2019) Optimal machine learning models for robust materials classification using ToF-SIMS data. Applied Surface Science 487: 773-783.
• Ma N, Liu S, Liu W, Xie L, Wei D, et al. (2020) Research Progress of Titanium-Based High Entropy Alloy: Methods, Properties, and Applications. Frontiers in Bioengineering and Biotechnology 8: 1303.
• Wu CT, Chang HT, Wu CY, Chen SW, Huang S Y, et al. (2020) Machine learning recommends affordable new Ti alloy with bone-like modulus. Materials Today 34: 41-50.
• Huang W, Martin P, Zhuang HL (2019) Machine-learning phase prediction of high- entropy alloys. Acta Materialia 169: 225-236.
• Wen C, Zhang Y, Wang C, Xue D, Bai Y, et al. (2019) Machine learning assisted design of high entropy alloys with desired property. Acta Materialia 170: 109-117.
• Zhang Y, Wen C, Wang C, Antonov S, Xue D, et al. (2020) Phase prediction in high entropy alloys with a rational selection of materials descriptors and machine learning models. Acta Materialia 185: 528-539.
• Concu R, Kleandrova VV, Speck-Planche A, Cordeiro MND (2017) Probing the toxicity of nanoparticles: a unified in silico machine learning model based on perturbation theory. Nanotoxicology 11(7): 891-906.
• Labouta HI, Asgarian N, Rinker K, Cramb DT (2019) Meta-analysis of nanoparticle cytotoxicity via data-mining the literature. ACS nano 13(2): 1583-1594.
• Lazarovits J, Sindhwani S, Tavares AJ, Zhang Y, Song F, et al. (2019) Supervised learning and mass spectrometry predicts the in vivo fate of nanomaterials. ACS nano 13(7): 8023-8034.
• Choi JS, Trinh TX, Yoon TH, Kim J, Byun HG (2019) Quasi-QSAR for predicting the cell viability of human lung and skin cells exposed to different metal oxide nanomaterials. Chemosphere 217: 243-249.
• Chen CT, Gu GX (2019) Effect of constituent materials on composite performance: exploring design strategies via machine learning. Advanced Theory and Simulations 2(6): 1900056.
• Han T, Stone-Weiss N, Huang J, Goel A, Kumar A (2020) Machine learning as a tool to design glasses with controlled dissolution for healthcare applications. Acta biomaterialia 107(15): 286-298.