Integrating SEResNet101 and SE-VGG19 for advanced cervical lesion detection: a step forward in precision oncology

Abstract

Background: Cervical cancer remains a significant global health issue, with accurate differentiation between low-grade (LSIL) and high-grade squamous intraepithelial lesions (HSIL) crucial for effective screening and management. Current methods, such as Pap smears and HPV testing, often fall short in sensitivity and specificity. Deep learning models hold the potential to enhance the accuracy of cervical cancer screening but require thorough evaluation to ascertain their practical utility.

Methods: This study compares the performance of two advanced deep learning models, SEResNet101 and SE-VGG19, in classifying cervical lesions using a dataset of 3,305 high-quality colposcopy images. We assessed the models based on their accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).

Results: The SEResNet101 model demonstrated superior performance over SE-VGG19 across all evaluated metrics. Specifically, SEResNet101 achieved a sensitivity of 95%, a specificity of 97%, and an AUC of 0.98, compared to 89% sensitivity, 93% specificity, and an AUC of 0.94 for SE-VGG19. These findings suggest that SEResNet101 could significantly reduce both over- and under-treatment rates by enhancing diagnostic precision.

Conclusion: Our results indicate that SEResNet101 offers a promising enhancement over existing screening methods, integrating advanced deep learning algorithms to significantly improve the precision of cervical lesion classification. This study advocates for the inclusion of SEResNet101 in clinical workflows to enhance cervical cancer screening protocols, thereby improving patient outcomes. Future work should focus on multicentric trials to validate these findings and facilitate widespread clinical adoption.

Keywords: Cervical cancer; Cervical screening; Deep learning; HSIL; Image classification; LSIL; Medical imaging; SE-VGG19; SEResNet101.

Fig. 1

The cervical Image Data Acquisition Process is normal with various Lesion Presentations

Fig. 2

Image Preprocessing Effects

Fig. 3

SEResNet101 Model Extracting Cervical ROI Effects

Fig. 4

Evaluation of Cervical Image Classification Performance by Different Models - Confusion Matrix, Specificity, and F1 Score Presentations. Note: (A, D) Confusion Matrix: The darker colors on the diagonal represent high accuracy in the classification of HSIL, LSIL, and normal cervical images. The closer the value is to 1, the higher the model’s prediction accuracy. (B, E) Specificity for each category: Line graphs depict the model’s specificity for each classification, particularly in accurately identifying true negatives, especially between HSIL and normal images compared to LSIL. (C, F) F1 scores for each category: Bar graphs display the F1 scores of HSIL, LSIL, and normal images, which are harmonic means of precision and recall, indicating the model’s balanced performance in correctly classifying each category

Fig. 5

Evaluation of Cervical Image Classification Performance by Different Models - ROC Curve and Precision-Recall Curve Representations. Note: (A, C) ROC Curve: Evaluates the classification’s effectiveness through the relationship between true positive rate and false positive rate. (B, D) Precision-Recall Curve: Provides the relationship between precision and recall, reflecting the model’s accuracy in positive class predictions and coverage

Fig. 6

Evaluation of cervical image classification performance by different models - display of training loss and learning rate variations. Note: (A, C) Training Loss: The chart illustrates the variation in training epochs, reflecting the model’s ability to optimize parameters and reduce prediction errors during the learning process. (B, D) Learning Rate Variation: Demonstrates the adjustment of the learning rate in each training epoch, which helps the model avoid local minima during training and aids in finding the global minimum of the loss function

Fig. 7

Evaluation of Cervical Image Classification Performance by Different Models - Display of mAP and Training Accuracy. Note: (A, C) mAP vs. Epoch: Showcases the increasing mean average precision (mAP) of two models in the cervical image classification task over training epochs, used to measure the overall performance improvement of the models. (B, D) Training Accuracy: Reflects the accuracy of the models over each training epoch, showing the model’s fitting capability on the training set over time

Fig. 8

Comprehensive Analysis Process and Major Conclusions of this Study