Cervical cancer is the second most common cancer that affects women after breast
cancer. This cancer can be detected at early stage with Pap test. At present, many
intelligent systems have been proposed to assist doctors to detect the cervical precancerous
stages. This project proposes an intelligent microcontroller based cervical
cancer diagnosis system. The Hybrid Multilayer Perceptron (HMLP) network, which is
trained using Modified Recursive Prediction Error (MRPE), is used as the intelligent
classifier to classify the cervical pre-cancer into three classes, namely normal, LSIL
(Low Grade Squamous Intraepithelial Lesion) and HSIL (High Grade Squamous
Intraepithelial Lesion). Four features of a cervical cell, namely nucleus size, cytoplasm
size, nucleus greylevel and cytoplasm greylevel are used as the input data. Two hundred
training data and one hundred testing data are used to test the performances of the
HMLP networks and thus obtain the optimum structures of the HMLP networks. These
HMLP networks with the obtained optimum structures will be implemented into the
microcontroller. Then, fifty data are used to test the proposed microcontroller system.
The results show that the proposed system produces good performances by having
percentages of accuracy, sensitivity and specificity which are 94.00%, 92.31% and
100.00% respectively. The obtained percentages of false negative and false positive are
7.69% and 0.00% respectively. These prove that the proposed microcontroller based
cervical pre-cancer diagnosis system is highly capable to be used as a good cervical precancer
diagnosis system.