Enhanced Cervical Cancer Classification Using Convolutional Tsetlin Machines with Transfer Learning

Cervical cancer remains a major global health concern, particularly in low-resource settings where access to skilled diagnostic personnel is limited. While Pap smear screening plays a vital role in early detection, manual analysis is time-consuming, error-prone, and subject to inter-observer variability. This study presents a novel hybrid model for cervical cancer classification that combines the deep feature extraction capabilities of the InceptionV3 convolutional neural network with the interpretable decision-making framework of Convolutional Tsetlin Machines (CTMs). Features extracted from Pap smear images are binarized and used as input for CTMs, which generate logical clauses to classify images into clinically relevant categories. Evaluated on a publicly available Pap smear dataset using 10-fold cross-validation, the proposed model achieved high performance with an average accuracy of 99.96%, precision of 98.99%, recall of 98.96%, and F1-score of 98.98%. These results suggest the model’s strong potential for clinical decision support, especially in settings where interpretability and transparency are critical. While promising, further evaluation on larger and more diverse datasets is necessary to validate generalizability and real-world applicability.