HSV-Driven Light-Weight CNN Architecture: Balancing Computational Complexity and Diagnostic Precision in Cervical Cancer Detection

Cervical cancer ranks among the foremost contributors to cancer-related mortality in women globally. Timely identification is essential for enhancing survival rates, and recent developments in deep learning have demonstrated potential in automating the categorization of cervical cancer cells. This paper presents a lightweight Convolutional Neural Network (CNN) model aimed at achieving efficient and accurate classification of cervical cancer images. The model utilizes HSV color space images, improving classification performance while eliminating the necessity for intricate preprocessing. The proposed architecture features a total of 74,629 parameters, ensuring computational efficiency without compromising on accuracy. The model was assessed using the SIPaKMeD dataset, resulting in an impressive accuracy of 97.34%. This performance exceeds that of various leading models, such as DenseNet-121 and VGG-16, which necessitate a considerably higher number of parameters. The model exhibits impressive precision, recall, and F1 scores, reflecting a well-rounded performance across various cervical cancer cell types. The lightweight CNN we developed effectively minimizes computational expenses while offering a viable approach for extensive screening and diagnosis of cervical cancer. The findings indicate that the proposed model could serve as an important asset for medical professionals in the early identification of cervical cancer, enhancing diagnostic results while utilizing minimal computational resources.