Diabetic Retinopathy (DR) is an eye disease that is the main cause of blindness in developed countries. Treatment of DR and prevention of blindness depend heavily on regular monitoring, early-stage diagnosis, and timely treatment. Vision loss can be effectively prevented by the automated diagnostic system that assists ophthalmologists who otherwise practice manual lesion detection processes which are tedious and time-consuming. Therefore, the purpose of this research is to design a system that can detect the presence of DR and be able to classify it based on its severity. In this proposed, the classification process is carried out based on image discovery by extracting GLCM texture features from 454 retinal fundus images in the IDRID database which are classified into 4 severity levels, namely normal, mild NPDR, moderate NPDR, and severe NPDR. The features obtained from each image will be used as input for the classification process using SVM. As a result, the classification system that has been trained is able to classify 4 levels of DR severity with an average accuracy of 89.55%, a sensitivity of 81.03%, and a specificity of 92.89%. Based on the results of the evaluation of the performance of this classification system, it can be concluded that the specificity value is higher than the specificity value, this indicates that the system that has been trained has a good ability to identify negative samples or those that indicate a class.

Low image resolution has deficiencies in the diagnostic process, this will affect the quality of the image in describing an object in certain tissues or organs, especially in the process of examining patients by doctors or physicians based on the results of imaging medical devices such as CT-scans, MRIs and X-rays. Therefore, this study had developed a General Regression Neural Network (GRNN) type artificial neural network system to reconstruct a medical image so that the image has a significant resolution for the analysis process. The GRNN input layer uses grayscale intensity values with variations in the image position coordinates to produce an optimal resolution. There are four layers in this method, the first is input layer, the second is hidden layer, the third is summation, and the last layer is output. We examined the two parameters with different interval values of 0.2 and of 0.5. The result shows that the interval value of 0.2 is the optimal value to produce an output image that is identical to the input image. This is also supported by the results of the intensity curve of the RGB pattern matched between target and output.

Pencitraan medis atau  medical imaging adalah  suatu cara untuk  mendapatkan informasi citra medis  tanpa   harus   menggunakan tindakan  operasi   atau   bedah.   Proses  diagnosis   dalam pencitraan medis akan memberikan informasi  terkait  bentuk,  lokasi, objek yang di teliti, atau disebut  dengan   ROI  (Region  of  Interest).  Pada   Penelitian   ini  dibuat   sebuah   rancangan metoda  segmentasi secara komputasi menggunakan teknik  watershed dengan  filter sobel dan morphological gradient untuk menganalisis daerah tumor dan mengurangi efek segmentasi berlebihan  yang  muncul  pada  teknik  watershed,  pada  citra  otak  dengan  tinjauan  tiga  slice hasil  citra  MRI  yang  berbeda  yaitu  axial,  koronal   dan  sagital.  Hasil  percobaan dari  dua metoda  kombinasi  teknik watershed makers dan morphological gradient menghasilkan segmentasi baik mengurangi  segmentasi yang berlebihan  serta hasil yang lebih tajam,  dengan hasil pengujian  kualitas  citra dengan  metoda  SNR  (Signal Noise to Ratio) untuk  setiap slice adalah  axial 5.73 dB, koronal  6.38 dB dan sagital 5.96 dB dengan waktu  rata-rata komputasi adalah  1.20 s dan kombinasi  segmentasi menggunakan filter sobel untuk  masing-masing  slice adalah  axial 5.68 dB, koronal  6.28 dB, dan sagital 5.27 dB dengan waktu rata-rata komputasi adalah1.80  s.