Medical X-ray images enhancement based on super resolution convolution neural network

Abstract

Pneumonia is a severe lung infection, chest X-ray (CXR) image preferred to find infection. Real images lost its quality, resolution and other feature due to transmission.So  good  qualitative  datasets  are  very  limited.Quality enhancement  in  medical  images  is  challenging  task  for  researchers.  And quality  in  clinical  diagnosis  of  any  disease  in  deep  learning  play  a  very important role.So,this paper presentsanaspect with importance of quality in  medical  images  CXR  of  a  particular  dataset  and  how  to  enhance  and create   new   images   with   high   quality   resolution,   that   is   re-used   for classification in deep learning.Super resolution convolutional neural netwok (SRCNN)  is  deep  learning  based  method,  which  is  used  for  improving resolution  in  image.  Super  resolution  means  low  resolution  (LR)  images from  dataset  is  to  be  reconstructed  or  magnified  into  high  resolution  (HR). The  objective  behind  this  study is  to  measure  the  effect of  super resolution with  quality  index, peak  signal-to-noise  ratio (PSNR), mean  squared  error (MSE),and structural  similarity  index  measure (SSIM).  This  experinment performed  on  200  images  with  10  batches,  each  batch  has  20  images  from Kermany  dataset,  select LRimages  and  converted  into HRwith  SRCNN. Then we find PSNR value of image is increase upto 2 to 5 DB, and MSE of qood  quality  images  is  near  to  zero  and  MSE  decrease  up  to  20-25,  SSIM value  have  little  variation  due  to  same  pattern  is  found  in  input  and  output images.  Enhancement  means  highlight  or  improve  the  region  of  interest  of pneumonic images. Main goal of this study is to preapare a modified dataset which is further used for classification.