之前一直使用别人的免费浏览工具来浏览DCM图像，或多或少都存在小的问题，要么完全免费但是功能不全不好用，要么就是收费需要定期下载版本申请试用，折腾来折腾去很是费心，决定最近自己写个简单的，不求功能强大只求自己用起来得心应手。
底层文件的读取使用DCMTK3.6.3的DCMData包；考虑支持跨平台，上层显示使用QT来做。
在Linux和windows两个平台下编译DCMTK生成的Config文件夹中的头文件内容是不同的，为了更好地组织两个平台的头文件和库文件，在dcmtk文件夹分别新建linux和win文件夹，将ubuntu下编译后的include文件夹拷贝到linux下，将win7下编译后的include文件夹拷贝到win下。读取文件主要使用DCMTK的DcmData库，该库依赖了ofstd,oflog库，为了支持RLE压缩和JPEG压缩，还需要dcmjpeg库。以linux为例，将编译后的config文件夹下的include目录和源码中的ofstd,oflog,dcmjpeg,dcmdata文件夹中的include文件夹拷贝到dcmtk/linux/include目录下，并在工程文件.pro中将这些目录添加到INCLUDEPATH中；
win32: INCLUDEPATH += $$PWD/dcmtk/win/include/ else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/ win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/config else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/config win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/ofstd else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/ofstd win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/ofstd/diag else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/ofstd/diag win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/ofstd/variadic else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/ofstd/variadic win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog/config else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog/config win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog/helpers else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog/helpers win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog/internal else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog/internal win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog/thread else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog/thread win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/oflog/spi else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/oflog/spi win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/dcmjpeg else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/dcmjpeg win32: INCLUDEPATH += $$PWD/dcmtk/win/include/dcmtk/dcmdata else:unix: INCLUDEPATH += $$PWD/dcmtk/linux/include/dcmtk/dcmdata
将这几个库文件也拷贝到dcmtk/linux/lib文件夹下，使用到的库文件和.pro代码如下：
DEPENDPATH += C:\Program Files\Microsoft SDKs\Windows\v6.0A\Lib\x64; win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lcharset_d else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lcharset_d else:unix: LIBS += -lcharset win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -llibiconv_d else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -llibiconv_d else:unix: LIBS += -liconv win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lofstd else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lofstd else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -lofstd win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -loflog else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -loflog else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -loflog win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -ldcmjpeg else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -ldcmjpeg else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -ldcmjpeg win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -ldcmimage else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -ldcmimage else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -ldcmimage win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -ldcmimgle else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -ldcmimgle else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -ldcmimgle win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lijg8 else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lijg8 else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -lijg8 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lijg12 else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lijg12 else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -lijg12 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lijg16 else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lijg16 else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -lijg16 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -ldcmdata else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -ldcmdata else:unix: LIBS += -L$$PWD/dcmtk/linux/lib/ -ldcmdata win32:CONFIG(release, debug|release): LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/x64/' -lNetAPI32 else:win32:CONFIG(debug, debug|release): LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/x64/' -lNetAPI32 win32: LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/x64/' -lWSock32 win32:CONFIG(release, debug|release): LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/xl64/' -lWS2_32 else:win32:CONFIG(debug, debug|release): LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/x64/' -lWS2_32 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/release/ -lzlib_d else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/dcmtk/win/lib/debug/ -lzlib_d unix:!macx: LIBS += -lz win32: LIBS += -L'C:/Program Files/Microsoft SDKs/Windows/v6.0A/Lib/' -lIPHlpApi win32: LIBS += -lAdvAPI32
需要添加的头文件如下：
#include "dcdeftag.h" #include "dcdatset.h" #include "dcelem.h" #include "dcfilefo.h" #include "dcuid.h" #include "dcrledrg.h" #include "dcmetinf.h" #include "djdecode.h"
接下来就可以使用DcmFielFormat类来读取图像信息了。
DcmFileFormat* m_pDcmFile; OFCondition result = dcmFile.loadFile(strFileName); if (result.bad()) { return false; } DcmDataset* dataset = dcmFile.getDataset(); if (dataset == NULL) { return false; }
首先，如果读取RLE或者JEPG压缩的图像，需要先使用DcmJpeg库来转换Transfer syntax，方法如下：
DcmMetaInfo* meta = dcmFile.getMetaInfo(); DcmElement *element = NULL; OFString transferSyntaxUID; result = meta->findAndGetElement(DCM_TransferSyntaxUID, element); if (result.bad() || element == NULL) { // assert(false); // return false; } else { element->getOFString(transferSyntaxUID, 0); } if (transferSyntaxUID.compare(UID_RLELosslessTransferSyntax)==0) { DcmRLEDecoderRegistration::registerCodecs(); result = dataset->chooseRepresentation(EXS_LittleEndianExplicit, NULL); DcmRLEDecoderRegistration::cleanup(); if (result.bad()) { return false; } } else if ( transferSyntaxUID.compare(UID_JPEGProcess14SV1TransferSyntax)==0 || transferSyntaxUID.compare(UID_JPEGProcess1TransferSyntax)==0 ) { DJDecoderRegistration::registerCodecs(); result = dataset->chooseRepresentation(EXS_LittleEndianExplicit, NULL); DJDecoderRegistration::cleanup(); if (result.bad()) { return false; } }
接下来读取像素相关的tag，主要读取C.6.3 Image Pixel Module也就是TableC.7-11a.Image Pixel Module Attributes的Tags。主要Tag的含义如下：
（0028,0002）Samples per Pixel：每个像素的存储单元个数，也就是几个存储单元数据来表示一个像素的信息。值为1或者3，其他值的含义没有定义，对于monochrome和palette color图像值为1，对于RGB或者其他vector color models，值为3.
（0028,0004）Photometric Interpretation:图像的类型，详见C7.6.3.1.2 Photometric Interpretation的解释，黑白灰度图像大多使用MONOCHROME1或者MONOCHROME2。
（0028,0010）Rows：图像的行数
（0028,0011）Columns：图像的列数
（0028,0100）Bits Allocated:每个Sample分配的bit数
（0028,0101）Bits Stored:每个Sample实际存储的bit数
（0028,0102）High Bit：每个sample的最高位
（0028，0103）Pixel Repersentation:Sample的数据表示形式，0表示无符号整型，1表示2的补码。
(7FE0,0010)Pixel Data：像素数据
（0028,0006）Planar Configuration:彩色像素数据的表示形式，是按像素来排布还是按颜色色素来排布，比如RGB图像，0表示RGBRGBRGB……RGB的形式来存储，1表示RRR……RRRGGG……GGGBBB……BBB的形式来存储。
（0028,0106）Smallest Image Pixel：本幅图中像素值的最小值
（0028,0107）Largest Image Pixel Value:本幅图像中像素值的最大值；
（0028,1101-1103）（0028,1201-1203）分别为RGB描述了一个查找表，如果图像是PALETTE COLOR，就利用像素值作为索引来查找这个表，从而得到真正的像素值。Descriptor包含三个数值，第一个值为查找表的元素个数；第二个值为最小的索引数，也就是（7FE0, 0010）中读取的最小数；第三个值为查找表的每一个元素的位数。
通过以上Tags就可以解析出来图像中每个像素的具体值，接下来要把这些像素值显示出来，一般需要经过两步转换，Modality LUT和VOI LUT。
Modality LUT: 将设备相关的像素值转换为设备无关的像素值，比如CT图像的Hounsfield units转换为与CT无关的像素值。如果是线性转换就使用Rescale Slope(0028,1053)和Rescale Intercept（0028,1052）来转换，如果是非线性就使用Modality LUT Sequnce(0028, 3000)来转换。
VOI LUT：在进行Modality LUT后，将得到的像素信息转换成显示像素信息，比如将像素信息缩放到可显示的范围内。如果是线性转换就通过Window Center(0028,1050)和Window Width(0028,1051)来转换，如果是非线性就通过VOI LUT Sequence(0028, 3010)来定义。在C11.2.1.2.2中明确提到只有MONO1和MONO2的图像才需要VOI LUT转换。
只要按照定义把像素值正确地读出来，再经过正确的LUT，一个完整的图像就得到了。
代码托管在github.com上，https://github.com/JorSean/dicom-explorer。欢迎搞DICOM开发的同仁们指点，大家一块学习一块进步。