DDSM数据库转换图像格式——LJPEG转为PNG格式

Digital Database for Screening Mammography (DDSM)数据库是一个非常大的乳腺图像数据库，有一万多张图像，但是图像格式是LJPEG，现有的图像软件（如photoshop、ACCDsee、windows自带的图像查看软件）以及编程软件（如matlab）都无法读取，需要将其转换成其他常见的格式才能使用。我从网上搜到了很多方法，试过之后都不成功，其中包括该数据库的创建者——南佛罗里达大学自己写的一个程序[1]，一个医学图像格式转换软件XMedCon[2]。最后成功的方法是使用曼彻斯特大学的Dr. Chris Rose写的一个完整的程序，在他的程序基础上做了些修改，成功的将图像格式转换成了PNG格式。他的程序链接见http://microserf.org.uk/academic/Software.html （PS. 最近发现此链接地址已失效，所以我把源程序放在了我的github上，地址：https://github.com/hd11224/DDSM，包括了所有需要的工具和软件）

 从上面给出的链接下载到的程序中，有用户手册告诉你怎么使用这个程序。这个程序是用Ruby语言写的，需要在Cygwin下运行，用户手册中有介绍如何安装Cygwin及其他需要的工具。程序工作流程是使用者手动输入图像名称，程序先从FTP上下载该图像，然后经过几步转换，最终转换为PNG格式。我在按照用户手册运行这个程序时没有成功，用VS2013打开get-ddsm-mammo文件查看源码，发现是从FTP上下载图像环节出了问题，做了修改后，最终运行成功。我修改后的程序如下： 

#!/usr/bin/ruby

# This program gets a specified mammogram from the DDSM website and
# converts it to a PNG image. See the help message for full details.

require 'net/ftp'


# Specify the name of the info-file.
def info_file_name
  'info-file.txt'
end

def image_names
  'image_name.txt'
end

# Get an FTP file as specified by a DDSM path (e.g.,
# /pub/DDSM/cases/cancers/cancer_06/case1141/A-1141-1.ics) and return the
# local path to the file, or return nil if the file could not be dowloaded.
def get_file_via_ftp(ddsm_path)
  ftp = Net::FTP.new('figment.csee.usf.edu')
  ftp.passive = true
  ftp.login  
  ftp.chdir(File.dirname(ddsm_path))  
  puts File.basename(ddsm_path)  
  ftp.getbinaryfile(File.basename(ddsm_path))
  #ftp.getbinaryfile(ddsm_path)
  
    # Will be stored local to this program, under the same file name
    
  # Check to make sure that we managed to get the file.
  if !FileTest.exist?(File.basename(ddsm_path))
    puts "Could not get the file #{File.basename(ddsm_path)} from the DDSM FTP server; perhaps the server is busy."
    exit(-1)
  end  
  return File.basename(ddsm_path)
end


# Return the string input with the system's filesep at the end; if there
# is one there already then return input.
def ensure_filesep_terminated(input)
  if input[input.length-1].chr != File::SEPARATOR
    input += File::SEPARATOR
  end

  return input
end

# Check program input; input is the program input (i.e ARGV).
def check_inputs(input)
  if input.length != 1
    puts get_help
    exit(-1)
  end
  
  # See if the user wanted the help docs.
  if input[0] == '--help'
      puts get_help
      exit(-1)
  end
  
  # Check to make sure that the info file exists.
  if !FileTest.exist?(info_file_name)
    puts "The file #{info_file_name} does not exist; use catalogue-ddsm-ftp-server.rb"
    exit(-1)
  end

end

# Given the name of a DDSM image, return the path to the
# .ics file associated with the image name. If we can't find the 
# path, then we return nil.
def get_ics_path_for_image(image_name)

  # Does image_name look right?
  if image_name[/._\d{4,4}_.\..+/].nil?
    raise 'image_name seems to be wrong. It is: ' + image_name
  end

  # Edit the image name, as .ics files have the format 'A-0384-1.ics';
  # there is no '.RIGHT_CC' (for example).
  image_name = image_name[0..(image_name.rindex('.')-1)] # Strip everything after and including the last '.'.
  image_name[1] = '-'
  image_name[6] = '-' # Change the '_'s to '-'s (better regexp-based approach?).
  image_name+='.ics' # Add the file suffix.

  # Get the path to the .ics file for the specified image.
  File.open(info_file_name) do |file|
    file.each_line do |line|
      # Does this line specify the .ics file for the specified image name?
      if !line[/.+#{image_name}/].nil?
        # If so, we can stop looking		
        return line		
      end
    end
  end
  
  # If we get here, then we did not find a match, so we will return nil.
  return nil
end

# Given a line from a .ics file, return a string that specifies the
# number of rows and cols in the image described by the line. The
# string would be '123 456' if the image has 123 rows and 456 cols.
def get_image_dims(line)
  rows = line[/.+LINES\s\d+/][/\d+/]
  cols = line[/.+PIXELS_PER_LINE\s\d+/][/PIXELS_PER_LINE\s\d+/][/\d+/]

  return rows + ' ' + cols
end

# Given an image name and a string representing the location of a
# local .ics file, get the image dimensions and digitizer name for
# image_name. Return a hash which :image_dims maps to a string of the
# image dims (which would be '123 456' if the image has 123 rows and
# 456 cols) and :digitizer maps to the digitizer name. If we can't
# determine the dimensions and/or digitizer name, the corresponding
# entry in the hash will be nil.
def get_image_dims_and_digitizer(image_name, ics_file)
  # Get the name of the image view (e.g. 'RIGHT_CC')
  image_view = image_name[image_name.rindex('.')+1..image_name.length-1]

  image_dims = nil
  digitizer = nil

  # Read the image dimensions and digitizer name from the file.
  File.open(ics_file, 'r') do |file|
    file.each_line do |line|
      if !line[/#{image_view}.+/].nil?
        # Read the image dimensions
        image_dims = get_image_dims(line)
      end
      if !line[/DIGITIZER.+/].nil?
        # Read the digitizer type from the file.
        digitizer = line.split[1].downcase # Get the second word in the DIGITIZER line.

        # There are two types of Howtek scanner and they are
        # distinguished by the first letter in image_name.
        if digitizer == 'howtek'
          if image_name[0..0].upcase == 'A'
            digitizer += '-mgh'
          elsif image_name[0..0].upcase == 'D'
            digitizer += '-ismd'
          else
            raise 'Error trying to determine Howtek digitizer variant.'
          end
        end
      end
    end
  end

  # Return an associative array specifying the image dimensions and
  # digitizer used.
  return {:image_dims => image_dims, :digitizer =>digitizer}
end

# Given the name of a DDSM image, return a string that describes
# the image dimensions and the name of the digitizer that was used to
# capture it. If 
def do_get_image_info(image_name)
  # Get the path to the ics file for image_name.
  ftp_path = get_ics_path_for_image(image_name)
  ftp_path.chomp!
  
  # Get the ics file; providing us with a string representing
  # the local location of the file.
  ics_file = get_file_via_ftp(ftp_path)

  # Get the image dimensions and digitizer for image_name.
  image_dims_and_digitizer = get_image_dims_and_digitizer(image_name, ics_file)

  # Remove the .ics file as we don't need it any more.
  File.delete(ics_file)

  return image_dims_and_digitizer
end



# Given a mammogram name and the path to the image info file, get the
# image dimensions and digitizer name string.
def get_image_info(image_name)
  # Get the image dimensions and digitizer type for the specified
  # image as a string.
  image_info = do_get_image_info(image_name)
  
  # Now output the result to standard output.
  all_ok = !image_info[:image_dims].nil? && !image_info[:digitizer].nil? # Is everything OK?
  if all_ok
    ret_val = image_info[:image_dims] + ' ' + image_info[:digitizer]
  end

  return ret_val
end

# Return a non-existant random filename.
def get_temp_filename
  rand_name = "#{rand(10000000)}" # A longish string
  if FileTest.exist?(rand_name)
    rand_name = get_temp_filename
  end

  return rand_name
end

# Retrieve the LJPEG file for the mammogram with the specified
# image_name, given the path to the info file. Return the path to the
# local file if successful. If we can't get the file, then return nil.
def get_ljpeg(image_name)
  # Get the path to the image file on the mirror of the FTP server.
  path = nil
  
  File.open(info_file_name) do |file|
    file.each_line do |line|
      if !line[/.+#{image_name}\.LJPEG/].nil?
        # We've found it, so get the file.
        line.chomp!
        local_path = get_file_via_ftp(line)		
        return local_path
      end
    end
  end

  # If we get here we didn't find where the file is on the server.
  return nil
end

# Given the path to the dir containing the jpeg program, the path to a
# LJPEG file, convert it to a PNM file. Return the path to the PNM
# file.
def ljpeg_to_pnm(ljpeg_file, dims_and_digitizer)
  # First convert it to raw format.
  command = "./jpeg.exe -d -s #{ljpeg_file}"
  `#{command}` # Run it.
  raw_file = ljpeg_file + '.1' # The jpeg program adds a .1 suffix.
  
  # See if the .1 file was created.
  if !FileTest.exist?(raw_file)
    raise 'Could not convert from LJPEG to raw.'
  end

  # Now convert the raw file to PNM and delete the raw file.
  command = "./ddsmraw2pnm.exe #{raw_file} #{dims_and_digitizer}"
  pnm_file = `#{command}`
  File.delete(raw_file)
  if $? != 0
    raise 'Could not convert from raw to PNM.'
  end

  # Return the path to the PNM file.
  return pnm_file.split[0]
end

# Convert a PNM file to a PNG file. pnm_file is the path to the pnm file
# and target_png_file is the name of the PNG file that we want created.
def pnm_to_png(pnm_file, target_png_file)
  command = "convert -depth 16 #{pnm_file} #{target_png_file}"
  `#{command}`

  if !FileTest.exist?(target_png_file)
    raise 'Could not convert from PNM to PNG.'
  end

  return target_png_file
end

#write image_names to image_nama.txt
def write_image_names(name)
  namefile=File.open(image_names,'a') 
  namefile.puts name
  namefile.puts "\r\n"
  namefile.close
end

# The entry point of the program.
def main  
  # Check to see if the input is sensible.
  #check_inputs(ARGV)
  
  #image_name = ARGV[0]

  File.open('read_names.txt','r') do |file|
    file.each_line do |line|
	  image_name = line
	  image_name.chomp!
	  
	  # Get the image dimensions and digitizer name string for the
	  # specified image.
	  image_info = get_image_info(image_name)
    
	  # Get the LJPEG file from the mirror of the FTP site, returning the
	  # path to the local file.
	  ljpeg_file = get_ljpeg(image_name)
    
	  # Convert the LJPEG file to PNM and delete the original LJPEG.
	  pnm_file = ljpeg_to_pnm(ljpeg_file, image_info)
	  File.delete(ljpeg_file)

	  # Now convert the PNM file to PNG and delete the PNG file.
	  target_png_file = image_name + '.png'
	  png_file = pnm_to_png(pnm_file, target_png_file)
	  File.delete(pnm_file)

	  # Test to see if we got something.
	  if !FileTest.exist?(png_file)
		raise 'Could not create PNG file.'
		exit(-1)
	  end

	  # Display the path to the file.
	  puts File.expand_path(png_file)

	  #write image name
	  write_image_names(image_name)

	  #exit(0)
	end  
     end
  exit(0)
end

# The help message
def get_help
  <

  (Note: the '\\' simply indicates that the above command should be on
  one line.)

  where:

  *  is the name of the DDSM image you want to get and
    convert, for example: 'A_1141_1.LEFT_MLO'.

  If successful, the program will print the path to the PNG file of
  the requested mammogram to standard output and will return a status
  code of 0. If unsuccessful, the program should display a
  useful error message and return a non-zero status code.

END_OF_HELP
end

# Call the entry point.
main

  很麻烦的一点是，原程序运行需要手动依次输入图像名称，一次只能处理一张图像，一张图像处理完后才能处理下一张，很费时费力，所以在上面贴出的程序中我还做了一点修改，可以批量处理图像。方法是将要处理的图像的名称提前写在一个txt文件里，一行一个，命名为read_names，运行程序只需输入 ./get-ddsm-mammo即可。程序运行界面如下：

 

    

  每处理完一张图像，程序会将图像的名称写在一个名为image_name的txt文件里，所以在运行程序前要先创建一个名为image_name的txt文件。

  最后一点要说明的是，用户手册中提到在安装Cygwin时要同时安装Ruby,因为当时的Cygwin版本较低，Ruby已不在手册中所示位置，而是单独拿出来的，要安装如下图所示的Ruby和rubygems：

和

  在ddsm-software文件夹下，会看到已经转为PNG格式的图像：

  

 

  [1]http://marathon.csee.usf.edu/Mammography/software/heathusf_v1.1.0.html

  [2]http://sourceforge.net/projects/xmedcon/