ENH: return sorted file list when reading DICOM series

The functions readImageDICOMFileSeries and readImageDICOMArrayBufferSeries
have an additional side effect that it sorts / orders the files
(particularly DICOM files) based on a pre-defined logic in gdcm.
However, this information is not related to the caller in any form.

Add Array<string> as the ordered list of filenames as additional output.

Author: jadh4v

src/core/Image.ts

@@ -13,7 +13,7 @@ class Image {
 
   size: number[]
 
-  metadata: Record<string, string | number | number[] | number[][]>
+  metadata: Record<string, string | string[] | number | number[] | number[][]>
 
   data: null | TypedArray
 

src/io/internal/pipelines/image/ReadDICOM/ReadImageDICOMFileSeries.cxx

@@ -30,6 +30,7 @@
 
 #include "itkPipeline.h"
 #include "itkOutputImage.h"
+#include "itkOutputTextStream.h"
 
 class CustomSerieHelper: public gdcm::SerieHelper
 {
@@ -194,14 +195,17 @@ int runPipeline(itk::wasm::Pipeline & pipeline, std::vector<std::string> & input
   OutputImageType outputImage;
   pipeline.add_option("-o,--output-image", outputImage, "Output image volume")->required();
 
+  itk::wasm::OutputTextStream outputFilenames;
+  auto outputFilenamesOption = pipeline.add_option("--output-filenames", outputFilenames, "Output sorted filenames.");
+
   ITK_WASM_PARSE(pipeline);
 
   typedef itk::QuickDICOMImageSeriesReader< ImageType > ReaderType;
   typename ReaderType::Pointer reader = ReaderType::New();
   reader->SetMetaDataDictionaryArrayUpdate(false);
 
   if (!singleSortedSeries)
-    {
+  {
     std::unique_ptr<CustomSerieHelper> serieHelper(new CustomSerieHelper());
     for (const std::string & fileName: inputFileNames)
     {
@@ -257,11 +261,21 @@ int runPipeline(itk::wasm::Pipeline & pipeline, std::vector<std::string> & input
     }
 
     reader->SetFileNames(fileNames);
-    }
+  }
   else
-    {
+  {
     reader->SetFileNames(inputFileNames);
+  }
+
+  // copy sorted filenames as additional output
+  if(!outputFilenamesOption->empty())
+  {
+    auto finalFileList = reader->GetFileNames();
+    for (auto f = finalFileList.begin(); f != finalFileList.end(); ++f)
+    {
+      outputFilenames.Get() << *f << ":";
     }
+  }
 
   auto gdcmImageIO = itk::GDCMImageIO::New();
   reader->SetImageIO(gdcmImageIO);
@@ -279,16 +293,25 @@ int main (int argc, char * argv[])
   std::vector<std::string> inputFileNames;
   pipeline.add_option("-i,--input-images", inputFileNames, "File names in the series")->required()->check(CLI::ExistingFile)->expected(1,-1);
 
+  // We are interested in reading --input-images beforehand.
+  // We need to add and then remove other options in order to do ITK_WASM_PARSE twice (once here in main, and then again in runPipeline)
   bool singleSortedSeries = false;
   auto sortedOption = pipeline.add_flag("-s,--single-sorted-series", singleSortedSeries, "There is a single sorted series in the files");
 
+  // Type is not important here, its just a dummy placeholder to be added and then removed.
   std::string outputImage;
   auto outputImageOption = pipeline.add_option("-o,--output-image", outputImage, "Output image volume")->required();
 
+  // Type is not important here, its just a dummy placeholder to be added and then removed.
+  std::string outputFilenames;
+  auto outputFilenamesOption = pipeline.add_option("--output-filenames", outputFilenames, "Output sorted filenames");
+
   ITK_WASM_PARSE(pipeline);
 
+  // Remove added dummy options. runPipeline will add the real options later.
   pipeline.remove_option(sortedOption);
   pipeline.remove_option(outputImageOption);
+  pipeline.remove_option(outputFilenamesOption);
 
   auto gdcmImageIO = itk::GDCMImageIO::New();
 

src/io/readImageDICOMArrayBufferSeries.ts

@@ -21,15 +21,16 @@ const workerFunction = async (
   )
   worker = usedWorker
 
-  const args = ['--memory-io', '--output-image', '0', '--input-images']
+  const args = ['--memory-io', '--output-image', '0', '--output-filenames', '1', '--input-images']
   fileDescriptions.forEach((desc) => {
     args.push(`./${desc.path}`)
   })
   if (singleSortedSeries) {
     args.push('--single-sorted-series')
   }
   const outputs = [
-    { type: InterfaceTypes.Image }
+    { type: InterfaceTypes.Image },
+    { type: InterfaceTypes.TextStream },
   ]
   const inputs = fileDescriptions.map((fd) => {
     return { type: InterfaceTypes.BinaryFile, data: fd }
@@ -54,6 +55,19 @@ const workerFunction = async (
     inputs
   }
   const result: PipelineResult = await webworkerPromise.postMessage(message, transferables)
+  const image: Image = result.outputs[0].data
+  const filenames: string[] = result.outputs[1].data.data.split(':')
+
+  if(image.metadata === undefined) {
+    const metadata:
+      Record<string, string | string[] | number | number[] | number[][]> = {}
+    metadata["orderedFileNames"] = filenames
+    image.metadata = metadata
+  }
+  else {
+    image.metadata["orderedFileNames"] = filenames
+  }
+
   return { image: result.outputs[0].data as Image, webWorker: worker }
 }
 const numberOfWorkers = typeof globalThis.navigator?.hardwareConcurrency === 'number' ? globalThis.navigator.hardwareConcurrency : 4
@@ -63,10 +77,12 @@ const seriesBlockSize = 8
 
 const readImageDICOMArrayBufferSeries = async (
   arrayBuffers: ArrayBuffer[],
-  singleSortedSeries = false
+  singleSortedSeries = false,
+  fileNames?: Array<string>
 ): Promise<ReadImageFileSeriesResult> => {
+  const validFileNames = fileNames && fileNames.length == arrayBuffers.length
   const fileDescriptions = arrayBuffers.map((ab, index) => {
-    return { path: `${index}.dcm`, data: new Uint8Array(ab) }
+    return { path: validFileNames? fileNames[index] : `${index}.dcm`, data: new Uint8Array(ab) }
   })
   if (singleSortedSeries) {
     const taskArgsArray = []

src/io/readImageDICOMFileSeries.ts

@@ -12,7 +12,8 @@ const readImageDICOMFileSeries = async (
   })
   const fileContents: ArrayBuffer[] = await Promise.all(fetchFileContents)
 
-  return await readImageDICOMArrayBufferSeries(fileContents, singleSortedSeries)
+  const fileNames = Array.from(fileList, (file) => file.name);
+  return await readImageDICOMArrayBufferSeries(fileContents, singleSortedSeries, fileNames)
 }
 
 export default readImageDICOMFileSeries