Deep Learning Based Histopathological Classification of Cervical Cancer Using YOLO-v8 and Inception-v3: A Comparative Performance Study

Usman Idris Ismail, Joshua C Shawulu, Mustapha Ismail, Albashir Ahmad, Sulaiman Yusuf Ali, Nuhu Abdulalim Muhammad, Saadatu Ali Jijji

Cervical Cancer Classification, Deep Learning, YOLO-v8, Inception-v3, Histopathology Images, Medical Image Analysis

Cervical cancer remains a major health concern among women worldwide, emphasizing the need for accurate and efficient diagnostic approaches. Deep learning has shown strong potential in automating medical image analysis and improving diagnostic reliability. This study aims to evaluate and compare the performance of two advanced deep learning models YOLO-v8 and Inception-v3 for the multi-class classification of cervical cancer histopathology images. A curated dataset of 225 histopathology images representing three cervical cancer subtypes (squamous cell carcinoma, adenocarcinoma, and adenosquamous carcinoma) was preprocessed using standardized resizing, normalization, and augmentation techniques. Both models were trained with optimized hyperparameters and evaluated using accuracy, precision, recall, F1-score, confusion matrices, and learning curve analysis to determine their effectiveness and generalization capability. The proposed workflow incorporates robust preprocessing, extensive augmentation, and systematic hyperparameter tuning to enhance model performance. YOLO-v8 leverages an efficient unified architecture for high-speed feature extraction, while Inception-v3 utilizes multi-scale convolutional processing to capture fine-grained morphological patterns within histopathology images. YOLO-v8 achieved an accuracy of 99.8% and Inception-v3 achieved 99.4%, demonstrating strong discriminative ability and reliable classification across all cancer subtypes. These results highlight the potential of deep learning models as effective tools for automated cervical cancer diagnosis. Despite the limited dataset size, the study provides a solid performance benchmark and establishes a foundation for future work incorporating larger datasets and multimodal diagnostic frameworks.

Received: 12 Sep 2025, Received in revised form: 25 Nov 2025, Accepted: 01 Dec 2025, Available online: 17 Dec 2025

10.22161/ijaers.1212.5