Deteksi tumor otak dan stroke hemoragik pada citra ct scan dengan analisis tekstur gray level co-occurrence matriks (GLCM)

TANSA, Salmawaty, Prof. Dr. Ir. Thomas Sri Widodo, DEA

2010 | Tesis | S2 Teknik Elektro

Saat ini ilmu kedokteran telah berkembang pesat, teknik diagnostic dan pengobatan telah memberikan harapan hidup bagi pasien. Salah satu factor yang mempengaruhi prognosa (harapan hidup) penderita tumor otak yaitu keunggulan teknik diagsnotik (CT Scan, MRI). Tumor otak dan stroke Hemoragik (pendarahan internal di otak) merupakan penyakit-penyakit berbahaya yang menyerang otak dan dapat mempengaruhi organ tubuh lainnya. Gambaran CT Scan tumor otak mirip dengan gambaran stroke hemoragik yang tampak seperti massa hyperdensity abnormal yang mendorong struktur otak sering membuat para dokter atau radiolog mengalami kesulitan dalam mendiagnosis. Dalam penelitian ini, akan menerapkan analisis tekstur menggunakan metode Gray Level Co-Occurrence Matrix(GLCM) pada 4 arah (0o, 45o, 90o, 135o) dengan parameter contrast, correlation, energy, homogenity untuk membedakan tekstur image tumor otak, stroke hemoragik dan normal sehingga menghasilkan nilai gold standard berdasarkan ciri-ciri tekstur yang ada. Pelatihan dan pengujian fitur-fitur tekstur menggunakan metode backpropagation jaringan syaraf tiruan dengan variasi nilai learning rate, dihasilkan pengujian yang terbaik pada learning rate 0,3 dengan persentase kesalahan sebesar 11%.

Today medical science has developed rapidly, diagnostic techniques and treatment have given hope of life for patients. One of the factors that affect prognosis (life expectancy) of brain tumor patients diagsnotik technical excellence (CT Scan, MRI). Brain tumors and hemorrhagic stroke (internal bleeding in the brain) is a dangerous disease that attacks the brain and can affect other organs. CT Scan of brain tumors similar to the picture of hemorrhagic stroke that looks like a mass of abnormal hyperdensity that encourages brain structure often makes the physician or radiologist had difficulty in diagnosing. In this study, will apply the texture analysis using Gray Level Cooccurrence Matrix (GLCM) in the fourth direction (0o, 45o, 90o, 135o) with the parameters of contrast, correlation, energy, homogenity to distinguish the texture image brain tumors, and hemorrhagic stroke normal so as to produce the gold standard values based on the characteristics of the existing texture Training and testing of texture features using neural networks with backpropagation learning rate variations, produced the best test on the learning rate by 0.3 percentage error is 11%.

Kata Kunci : CT,Scan,Tumor otak dan stroke,Analisis tekstur gray level co,occurrence matriks, brain tumors, hemorrhagic stroke, gray level co-occurrence matrix