On Machine Learning Analysis of Atomic Force Microscopy Images for Image Classification - sample surface recognition

Aug 4, 2026

Upload by Ophelia

Atomic force microscopy (AFM) imaging provides high-resolution digital maps of sample-surface topography together with multiple physicochemical properties, making it well suited for machine learning (ML) analysis in image classification and recognition. Limited acquisition speed and a small available dataset constrain broad use of deep learning, so the work focuses on ML strategies beyond popular convolutional neural networks. A template tailored to AFM is proposed, with a worked example for identifying cell phenotype, emphasizing statistical significance and offering a practical method to assess it.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Prospective ML recognition/classification for AFM

Why AFM matches ML image analysis

Challenges: slow imaging and noise

AFM control vs image classification

Limits of deep learning and small databases

ML template and cell phenotype example

Statistical significance and assessment method

On machine learning analysis of atomic force microscopy images for image classification, sample surface recognition I. Sokolov 1,2,3 * 1. Department of Mechanical Engineering, Tufts University, Medford, MA 02155, USA 2 Department of Biomedical Engineering, Tufts University, Medford, MA 02155, USA 3 Department of Physics, Tufts University, Medford, MA 02155, USA * Email: [Igor.Sokolov@Tufts.edu](Igor.Sokolov@Tufts.edu) Atomic force microscopy (AFM or SPM) imaging is one of the best matches with machine learning (ML) analysis among microscopy techniques. The digital format of AFM images allows for direct utilization in ML algorithms without the need for additional processing. Additionally, AFM enables the simultaneous imaging of distributions of over a dozen different physicochemical properties of sample surfaces, a process known as multidimensional imaging. While this wealth of information can be challenging to analyze using traditional methods, ML provides a seamless approach to this task. However, the relatively slow speed of AFM imaging poses a challenge in applying deep learning methods broadly used in image recognition. This Prospective is focused on ML recognition/classification when using a relatively small number of AFM images, small database. We discuss ML methods other than popular deep-learning neural networks. The described approach has already been successfully used to analyze and classify the surfaces of biological cells. It can be applied to recognize medical images, specific material processing, in forensic studies, even to identify the authenticity of arts. A general template for ML analysis specific to AFM is suggested, with a specific example of the identification of cell phenotype. Special attention is given to the analysis of the statistical significance of the obtained results, an important feature that is often overlooked in papers dealing with machine learning. A simple method for finding statistical significance is also described. Artificial Intelligence (AI) and its central part, Machine Learning (ML), are proliferating in all areas of research and development. ML has already been introduced to microscopy. 1-3 It has been demonstrated for the electron, 4, 5 Raman, 6, 7 optical, 8, 9 X-ray microscopies. 2, 10 The importance of image recognition is hard to overestimate. It can be used to make medical diagnoses, to predict effective treatments (prognostics), identify the origin of artifacts, etc. The use of ML analysis of AFM images has already been demonstrated to detect cancer 11, 12, to identify different cell phenotypes at the level of single cells. 13 The AFM technique is fundamentally different from the other imaging methods. 14-18 AFM allows not only imaging a sample surface but also obtaining a large number of physical and chemical parameters of the sample surface. 19-21 AFM allows for attaining a very high spatial resolution. Its high resolution comes with the price; AFM is extremely sensitive to multiple noises. This may result in hard-to-identify artifacts. 22 These and other factors are paramount when combining AFM with ML. ML analysis has also been used to control image acquisition in AFM, 17, 23 to improve the accuracy of nanomechanical measurements, 24, 25 to help do the reconstruction of sample properties and structures that are difficult to find using classical mathematical methods, 25-29 and to control imaging. 23, 30-32 These applications are typically specific to particular models and samples. For example, the AFM control depends on the intermolecular and interatomic forces acting between the AFM probe and sample. The variety of these forces, their combinations, and the environments in which the sample can be immersed is too high to make a universal feedback control based on ML. However, ML can simplify and even fully automate the AFM operation for a particular type of sample. 33 There were a few attempts to use multichannel information that can be recorded at each pi

FAQs

Why are AFM images suitable for machine learning-based classification?

AFM imaging outputs digital images and enables extraction of many physicochemical parameters of a sample surface, providing rich input for ML algorithms.

What main obstacle limits applying deep learning to AFM image recognition?

AFM imaging is relatively slow, so generating a large enough image database for convolutional neural networks is difficult.

What does the proposed approach emphasize beyond accuracy?

It highlights analyzing the statistical significance of obtained results, and it also describes a simple method for determining statistical significance.

Aug 4, 2026

Upload by Ophelia