Tuberkulosis merupakan penyakit menular dan merupakan salah satu penyebab kematian utama penduduk dunia umumnya, dan penduduk indonesia khususnya. Diagnosis tuberkulosis dilakukan melalui pemeriksaan gejala klinis, fisik/jasmani, pemeriksaan bakteriologis, radiologis dan pemeriksaan penunjang lain. Pemeriksaan radiologis dan pengetahuan untuk menilai suatu x-ray menyebabkan pemeriksaan x-ray merupakan keharusan rutin, sehingga tanpa pemeriksaan x-ray dianggap kurang lengkap. Berdasarkan hasil pemeriksaan x-ray, dapat ditemukan ciri-ciri tuberkulosis. Ekstraksi ciri merupakan bagian yang sangat penting dalam menggali ciri suatu citra x-ray. Beberapa metode statistis yang digunakan dalam mencari ciri seperti histogram, Gray level Co-occurence Matrix (GLCM), dan Principle Component Analysis (PCA) belum menghasilkan akurasi klasifikasi terbaik. Memperbaiki akurasi klasifikasi dapat dilakukan dengan menentukan pemilihan fitur terbaik. Untuk mendapatkan fitur terbaik, dapat dilakukan dengan mengkombinasi beberapa fitur yang dihasilkan oleh beberapa metode ekstraksi ciri. Dalam penelitian ini akan diterapkan metode kombinasi ekstraksi ciri berbasis statistis, yaitu pencarian fitur ciri citra x-ray paru diagnosis tuberkulosis berdasar histogram dan GLCM. Fitur ciri kombinasi histogram GLCM antara lain variance, std deviasi, skewness, kurtosis, contrast dan energy. Hasil penelitian menunjukkan bahwa metode kombinasi histogram GLCM, lebih baik dari metode histogram, metode GLCM, dan metode PCA. Hasil pengujian terhadap 33 citra uji yang diklasifikasikan kedalam 4 kelompok citra PA (Postero-Anterior) normal, PA abnormal, AP (Antero-Posterior) normal, AP Abnormal, diperoleh akurasi klasifikasi citra berdasar histogram (81,81%), GLCM (96,96%), PCA (81,82%), dan kombinasi histogram GLCM sebesar 100%. Oleh karena itu, metode kombinasi histogram GLCM merupakan metode yang tepat dalam menghasilkan fitur ciri terbaik dari citra x-ray paru diagnosis tuberkulosis.

Tuberculosis is an infectious disease and is one of the leading causes of death world population and the general population, especially Indonesia. Diagnosis of tuberculosis is done through the examination of clinical symptoms, physical, bacteriological examination, radiological and other investigations. Examination and radiological examination to assess the knowledge of an x-ray examination of x-ray causes a routine necessity, so that no x-ray examination is considered incomplete. Based on the results of an x-ray examination, can be found the characteristics of the image of tuberculosis. Feature extraction is a very important part in exploring the characteristics of an x-ray image. Several statistical methods were used in the search for characteristics such as histogram, Gray level co-occurence Matrix (GLCM), and Principle Component analysis (PCA) has not produced the best classification accuracy. Improving the accuracy of classification can be done by determining the selection of the best features. To get the best features, can be done by combining several features produced by several methods of feature extraction. In this study applied a method based on a statistical combination of feature extraction, namely the search for characteristic features of x-ray image diagnosis pulmonary tuberculosis based on histogram and GLCM. Features characteristic combination histogram GLCM is variance, std deviation, skewness, kurtosis, contrast and energy. The results showed that the combination method histogram GLCM, better than histogram, GLCM, and PCA methods. The test results of 33 test images were classified into 4 groups of images PA (postero-anterior) normal, PA abnormal, AP (Antero-Posterior) normal, AP abnormal, obtained image classification accuracy based on histogram (81,81%), GLCM (96,96%), PCA (81,82%), and the combination histogram GLCM (100%). Therefore, combination method histogram GLCM is an appropriate method to produce the best characteristic feature of the x-ray image diagnosis of pulmonary tuberculosis.

Kata Kunci : Ekstraksi Ciri, Citra X-Ray, Statistis, Histogram, GLCM, PCA