Perkembangan teknologi yang semakin pesat dewasa kini sedikit sebanyak telah memberi sumbangan dalam bidang sains dan teknologi. Kemajuan yang dicapai dalam aspek penyelidikan perubatan dapat membantu mengesan jangkitan kanser yang semakin menular. Pengesanan kanser pada peringkat awal sememangnya amat penting bagi membolehkan rawatan dibuat. Penemuan baru kaedah rangkaian neural yang semakin popular dan aplikasinya dalam bidang perubatan dapat memudahkan proses pengesanan sel-sel kanser. Bukan itu sahaja, malah kebolehharapan sesetengah kaedah ini juga telah dilaporkan lebih tepat berbanding dengan kaedah konvensional. Justeru itu, perlaksanaan projek ini, adalah bertujuan untuk mengkaji keberkesanan kaedah rangkaian neural dalam mengklasifikasi sampel-sampel sel kanser payudara. Dalam projek ini, klasifikasi sel-sel kanser payudara telah dilaksanakan berdasarkan kaedah Peta-peta Penubuhan-Diri Kohonen 2-Dimensi. Lima masukan dan satu keluaran telah dikenakan bagi menentukan sama ada sel telah dijangkiti kanser atau tidak. Satu set data latihan telah digunakan bagi membolehkan rangkaian mempelajari melakukan pengesanan sel-sel kanser payudara. Algoritma sanggaan belakang telah dipilih bagi melatih rangkaian ini. Satu set data ujian juga telah digunakan untuk menentusahkan prestasi model rangkaian ini. Hasil daripada penyelidikan yang dibuat, keputusan telah menunjukkan bahawa penggunaan kaedah rangkaian neural ini, mampu menghasilkan peratus ketepatan diagnosis setinggi 96.53% bagi set data latihan dan 92.75% bagi set data ujian. Tiga ciri dominan sel payudara yang telah dapat dikenalpasti adalah sel terpisah, panjang X dan panjang Y. Projek ini telah dilaksanakan dengan menggunakan perisian ‘NeuralWorks Professional II / Plus’.
_________________________________________________________________________________________
The rapid era of technology advancement has contributed a big impact against science and technology approach. This development achievement in medical investigation could helps to detect cancer infection which expanded mostly among the women. Cancer detection in its earliest stage definitely is very important for an effective treatment. New innovation in neural network methods which have become popular and its application in medical field have enabled prediction of the cancer cells easier. Not just that, the ability some of those methods have also been reported more accurate as compared to conventional methods. Therefore, the purpose of this project is to investigate the possibility of neural network methods in classifying breast cancer cells. In this project, breast cancer cells classification was achieved based on Self-Organizing Maps Kohonen 2-Dimension neural network. Five inputs and an output were used to detect whether the cells are cancerous or not. A set of training data was used to train the network. Back Propagation algorithm has been chosen to train the network. A set of testing data was used to determine the potential of this network model. The results showed that the application of this neural network method was able to achieve 96.53% of classification accuracy for training set and 92.75% for testing set. Discrete cell, X length and Y length were the three dominant features that have been identified. This project was implemented using ‘NeuralWorks Professional II / Plus’ software.
.
Klasifikasi sel-sel kanser payudara menggunakan rangkaian neural peta-peta penubuhan-diri (som)_Nuryanti Mohd Salleh_E3_2006_NI