Machine Learning Allows for Distinguishing Precancerous and Cancerous Human Epithelial Cervical Cells Using High-Resolution AFM Imaging of Adhesion Maps - Article

Q: Why is early detection of cervical cancer emphasized in this study?

Early detection substantially decreases morbidity and mortality associated with cervical cancer. The document frames the search for new early-diagnosis approaches as clinically significant.

Q: How does the proposed method improve over using only fractal dimension from AFM images?

The study reports that machine learning on high-resolution adhesion maps yields better separation between precancerous and cancerous cells, improving AUC and sensitivity versus the prior fractal-dimension-only approach.

Q: What algorithm and validation strategy are used to assess classification robustness?

A random forest decision tree algorithm is used. Robustness is verified with K-fold cross-validation (K = 500) and statistical significance is evaluated using a random shuffle control with p < 0.0001.

Sep 24, 2026

Uploaded by Ben

High-resolution adhesion maps from atomic force microscopy (AFM) are used to distinguish precancerous and cancerous human epithelial cervical cells. Compared with prior approaches relying only on fractal dimension, machine learning improves discrimination performance substantially, with reported AUC 0.93 versus 0.79 and higher sensitivity (92% vs 58%). A random forest decision tree model analyzes adhesion-map data from six precancerous and six cancerous primary cell lines. Robustness is validated via K-fold cross-validation (K = 500) and statistical significance is assessed with p < 0.0001 using a random shuffle control.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Introduction

Clinical need for early cervical cancer detection

Screening approaches and limitations of Pap smear and DNA-HPV tests

Methods

Machine learning using adhesion maps

Random forest model and cross-validation

Results

Classification performance for precancerous vs cancerous cells

Discussion

Sensitivity improvements and clinical implications

cells Article Machine Learning Allows for Distinguishing Precancerous and Cancerous Human Epithelial Cervical Cells Using High-Resolution AFM Imaging of Adhesion Maps Mikhail Petrov 1 and Igor Sokolov 1,2, * Citation: Petrov, M.; Sokolov, I. Machine Learning Allows for Distinguishing Precancerous and Cancerous Human Epithelial Cervical Cells Using High-Resolution AFM Imaging of Adhesion Maps. Cells 2023, 12, 2536. [https://doi.org/](https://doi.org/)[ ](https://doi.org/)[10.3390/cells12212536](10.3390/cells12212536) Academic Editors: Yuri M. Efremovand Petr V. Gorelkin Received: 22 September 2023 Revised: 19 October 2023 Accepted: 25 October 2023 Published: 28 October 2023 Copyright: © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 Department of Mechanical Engineering, Tufts University, Medford, MA 02155, USA; [mikhail.petrov@tufts.edu](mikhail.petrov@tufts.edu) 2 Departments of Physics and Biomedical Engineering, Tufts University, Medford, MA 02155, USA * Correspondence: [igor.sokolov@tufts.edu](igor.sokolov@tufts.edu) Abstract: Previously, the analysis of atomic force microscopy (AFM) images allowed us to distinguish normal from cancerous/precancerous human epithelial cervical cells using only the fractal dimension parameter. High-resolution maps of adhesion between the AFM probe and the cell surface were used in that study. However, the separation of cancerous and precancerous cells was rather poor (the area under the curve (AUC) was only 0.79, whereas the accuracy, sensitivity, and speciﬁcity were 74%, 58%, and 84%, respectively) . At the same time, the separation between premalignant and malignant cells is the most signiﬁcant from a clinical point of view. Here, we show that the introduction of machine learning methods for the analysis of adhesion maps allows us to distinguish precancerous and cancerous cervical cells with rather good precision (AUC, accuracy, sensitivity, and speciﬁcity are 0.93, 83%, 92%, and 78%, respectively) . Substantial improvement in sensitivity is signiﬁcant because of the unmet need in clinical practice to improve the screening of cervical cancer (a relatively low speciﬁcity can be compensated by combining this approach with other currently existing screening methods) . The random forest decision tree algorithm was utilized in this study. The analysis was carried out using the data of six precancerous primary cell lines and six cancerous primary cell lines, each derived from different humans. The robustness of the classiﬁcation was veriﬁed using K-fold cross-validation (K = 500) . The results are statistically signiﬁcant at p < 0.0001 . Statistical signiﬁcance was determined using the random shufﬂe method as a control. Keywords: cancer detection; machine learning; atomic force microscopy (AFM); scanning probe microscopy (SPM) 1. Introduction Morbidity and mortality associated with cervical cancer are substantially decreased when cancer is detected early [1–3] . Thus, the search for new approaches to early diagnosis is of great signiﬁcance. Cervical cancer is the second most frequent type of cancer among women worldwide, with approximately 288,000 deaths per year; more than 14,000 women were diagnosed with this disease in 2022 in the US alone [4,5] . The mortality rate is second only to that for breast cancer. The early detection of cervical cancer using the Papanicolaou (Pap) smear test has decreased mortality from cervical cancer by 70–80%[4,6] . Early detection is based on the identiﬁcation of neoplastic cells in stained preparations obtained from the uterine cervix [2,6] . It should be noted that, currently, recently introduced DNA-HPV tests are broadly used. I

FAQs

Why is early detection of cervical cancer emphasized in this study?

Early detection substantially decreases morbidity and mortality associated with cervical cancer. The document frames the search for new early-diagnosis approaches as clinically significant.

How does the proposed method improve over using only fractal dimension from AFM images?

The study reports that machine learning on high-resolution adhesion maps yields better separation between precancerous and cancerous cells, improving AUC and sensitivity versus the prior fractal-dimension-only approach.

What algorithm and validation strategy are used to assess classification robustness?

A random forest decision tree algorithm is used. Robustness is verified with K-fold cross-validation (K = 500) and statistical significance is evaluated using a random shuffle control with p < 0.0001.

Sep 24, 2026

Uploaded by Ben