Establishment and assessment of an early screening model for cervical cancer based on single-cell Raman spectroscopy combined with machine learning algorithms
Objective To establish an early screening model for cervical cancer based on single-cell Raman spectroscopy (SCRS) combined with machine learning algorithms, and to assess the performance of the model. Methods Cervical exfoliated cell samples were collected from 128 patients who were treated in our...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Journal of Precision Medicine
2025-08-01
|
| Series: | 精准医学杂志 |
| Subjects: | |
| Online Access: | https://jpmed.qdu.edu.cn/fileup/2096-529X/PDF/1754471584044-1741177256.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Objective To establish an early screening model for cervical cancer based on single-cell Raman spectroscopy (SCRS) combined with machine learning algorithms, and to assess the performance of the model. Methods Cervical exfoliated cell samples were collected from 128 patients who were treated in our hospital from September 2023 to June 2024, among whom 65 had normal results of ThinPrep cytologic test (TCT), 35 had abnormal TCT results, and 28 did not receive TCT. R language was used to divide the 100 cervical exfoliated cell samples with TCT results into training set and test set at a ratio of 8∶2, and SCRS was performed for all samples. Based on the SCRS data of the training set, 7 machine learning algorithms (KNN, PLS, LDA, RF, SVM, SVMRBF, and Stack) were used to establish an early screening model for cervical cancer, which was applied in the test set to identify the optimal model. The optimal model was then used to predict the TCT results of 100 cervical exfoliated cell samples in the training and test sets, which were compared with the actual TCT results. The remaining 28 samples without prior TCT results were used as a validation set and were subjected to TCT, and the optimal model was used to predict the TCT results of these samples, which were compared with the actual TCT results. Results There were significant differences in the relative intensities of characteristic Raman peaks at 874, 935, 1 024, 1 119, 1 250, 1 328, 1 569, and 1 642 cm-1 between the cervical exfoliated cells negative for intraepithelial lesion or malignancy, atypical squamous cells of undetermined significance, and the cervical exfoliated cells of low-grade squamous intraepithelial lesion. Among the 7 algorithms, the stacking model showed the best performance, with an AUC of 0.987, an accuracy of 99.2%, a sensitivity of 98.9%, and a specificity of 99.3%. In both training and test sets, the results predicted by the Stack model were relatively highly consistent with actual TCT results, with an accuracy of 91.0%, a sensitivity of 91.0%, a specificity of 87.4%, and an F1-score of 90.3%. In the validation set, the Stack model achieved an accuracy of 96.4%, a sensitivity of 100.0%, a specificity of 95.5%, and an F1-score of 92.3% in predicting TCT results. Conclusion The early screening model for cervical cancer based on SCRS and machine learning algorithms has a good performance and can be used as a noninvasive, efficient, and rapid tool to facilitate the early screening of cervical cancer. |
|---|---|
| ISSN: | 2096-529X |