Merge pull request #142 from sustrev/main

E2E dicom and metadata updates

oct_converter/dicom/dicom.py

@@ -122,6 +122,8 @@ def populate_opt_series(ds: Dataset, meta: DicomMetadata) -> Dataset:
     ds.StudyInstanceUID = generate_uid()
     ds.SeriesInstanceUID = generate_uid()
     ds.Laterality = meta.series_info.laterality
+    ds.ProtocolName = meta.series_info.protocol
+    ds.SeriesDescription = meta.series_info.description
     # Ophthalmic Tomography Series PS3.3 C.8.17.6
     ds.Modality = "OPT"
     ds.SeriesNumber = int(meta.series_info.series_id)
@@ -201,7 +203,9 @@ def write_opt_dicom(
     ds.ImageType = ["DERIVED", "SECONDARY"]
     ds.SamplesPerPixel = 1
     if meta.series_info.acquisition_date:
-        ds.AcquisitionDateTime =  meta.series_info.acquisition_date.strftime("%Y%m%d%H%M%S.%f")
+        ds.AcquisitionDateTime = meta.series_info.acquisition_date.strftime(
+            "%Y%m%d%H%M%S.%f"
+        )
     else:
         ds.AcquisitionDateTime = ""
 
@@ -269,10 +273,19 @@ def write_fundus_dicom(
     ds = populate_opt_series(ds, meta)
     ds.Modality = "OP"
     ds = populate_ocular_region(ds, meta)
-    ds = opt_shared_functional_groups(ds, meta)
+
+    ds.PixelSpacing = meta.image_geometry.pixel_spacing
+    ds.ImageOrientationPatient = meta.image_geometry.image_orientation
 
     # OPT Image Module PS3.3 C.8.17.7
     ds.ImageType = ["DERIVED", "SECONDARY"]
+    enface_to_type = {
+        "IR": "RED",
+        "FA": "BLUE",
+        "ICGA": "GREEN",
+    }
+    if ds.ProtocolName in enface_to_type:
+        ds.ImageType.append(enface_to_type.get(ds.ProtocolName))
     ds.SamplesPerPixel = 1
     ds.AcquisitionDateTime = (
         meta.series_info.acquisition_date.strftime("%Y%m%d%H%M%S.%f")
@@ -323,10 +336,19 @@ def write_color_fundus_dicom(
     ds = populate_opt_series(ds, meta)
     ds.Modality = "OP"
     ds = populate_ocular_region(ds, meta)
-    ds = opt_shared_functional_groups(ds, meta)
+
+    ds.PixelSpacing = meta.image_geometry.pixel_spacing
+    ds.ImageOrientationPatient = meta.image_geometry.image_orientation
 
     # OPT Image Module PS3.3 C.8.17.7
     ds.ImageType = ["DERIVED", "SECONDARY"]
+    enface_to_type = {
+        "IR": "RED",
+        "FA": "BLUE",
+        "ICGA": "GREEN",
+    }
+    if ds.ProtocolName in enface_to_type:
+        ds.ImageType.append(enface_to_type.get(ds.ProtocolName))
     ds.SamplesPerPixel = 1
     ds.AcquisitionDateTime = (
         meta.series_info.acquisition_date.strftime("%Y%m%d%H%M%S.%f")
@@ -368,6 +390,8 @@ def create_dicom_from_oct(
     interlaced: bool = False,
     diskbuffered: bool = False,
     extract_scan_repeats: bool = False,
+    scalex: float = 0.01,
+    slice_thickness: float = 0.05,
 ) -> list:
     """Creates a DICOM file with the data parsed from
     the input file.
@@ -382,6 +406,8 @@ def create_dicom_from_oct(
             interlaced: If .img file, allows for setting interlaced
             diskbuffered: If Bioptigen .OCT, allows for setting diskbuffered
             extract_scan_repeats: If .e2e file, allows for extracting all scan repeats
+            scalex: If .e2e file, allows for manually setting x scale (in mm)
+            slice_thickness: If .e2e file, allows for manually setting z scale (in mm)
 
     Returns:
             list: list of Path(s) to DICOM file
@@ -410,7 +436,13 @@ def create_dicom_from_oct(
             # if BOCT raises, treat as POCT
             files = create_dicom_from_poct(input_file, output_dir)
     elif file_suffix == "e2e":
-        files = create_dicom_from_e2e(input_file, output_dir, extract_scan_repeats)
+        files = create_dicom_from_e2e(
+            input_file,
+            output_dir,
+            extract_scan_repeats,
+            scalex,
+            slice_thickness,
+        )
     else:
         raise TypeError(
             f"DICOM conversion for {file_suffix} is not supported. "
@@ -471,7 +503,11 @@ def create_dicom_from_boct(
 
 
 def create_dicom_from_e2e(
-    input_file: str, output_dir: str = None, extract_scan_repeats: bool = False
+    input_file: str,
+    output_dir: str = None,
+    extract_scan_repeats: bool = False,
+    scalex: float = 0.01,
+    slice_thickness: float = 0.05,
 ) -> list:
     """Creates DICOM file(s) with the data parsed from
     the input file.
@@ -480,13 +516,17 @@ def create_dicom_from_e2e(
             input_file: E2E file with OCT data
             output_dir: Output directory
             extract_scan_repeats: If True, will extract all scan repeats
+            scalex: Manually set scale of x axis
+            slice_thickness: Manually set scale of z axis
 
     Returns:
             list: List of path(s) to DICOM file(s)
     """
     e2e = E2E(input_file)
-    oct_volumes = e2e.read_oct_volume()
-    fundus_images = e2e.read_fundus_image(extract_scan_repeats=extract_scan_repeats)
+    oct_volumes = e2e.read_oct_volume(scalex=scalex, slice_thickness=slice_thickness)
+    fundus_images = e2e.read_fundus_image(
+        extract_scan_repeats=extract_scan_repeats, scalex=scalex
+    )
     if len(oct_volumes) == 0 and len(fundus_images) == 0:
         raise ValueError("No OCT volumes or fundus images found in e2e input file.")
 

oct_converter/dicom/e2e_meta.py

@@ -37,13 +37,14 @@ def e2e_patient_meta(meta: dict) -> PatientMeta:
     return patient
 
 
-def e2e_series_meta(id, laterality, acquisition_date) -> SeriesMeta:
+def e2e_series_meta(id, laterality, acquisition_date, metadata) -> SeriesMeta:
     """Creates SeriesMeta from info parsed by the E2E reader
 
     Args:
         id: Equivalent to oct.volume_id or fundus.image_id
         laterality: R or L, from image.laterality
         acquisition_date: Scan date for OCT, or None for fundus
+        metadata: Additional metadata
     Returns:
         SeriesMeta: Series metadata populated by oct
     """
@@ -56,6 +57,16 @@ def e2e_series_meta(id, laterality, acquisition_date) -> SeriesMeta:
     series.laterality = laterality
     series.acquisition_date = acquisition_date
     series.opt_anatomy = OPTAnatomyStructure.Retina
+    if metadata.get("examined_structure", {}).get(id):
+        structure = metadata["examined_structure"][id]
+        try:
+            series.opt_anatomy = getattr(OPTAnatomyStructure, structure)
+        except AttributeError:
+            series.opt_anatomy = OPTAnatomyStructure.Unspecified
+    if metadata.get("enface_modality", {}).get(id):
+        series.protocol = metadata["enface_modality"][id]
+    if metadata.get("scan_pattern", {}).get(id):
+        series.description = metadata["scan_pattern"][id]
 
     return series
 
@@ -88,7 +99,8 @@ def e2e_image_geom(pixel_spacing: list) -> ImageGeometry:
     """
     image_geom = ImageGeometry()
     image_geom.pixel_spacing = [pixel_spacing[1], pixel_spacing[0]]
-    image_geom.slice_thickness = pixel_spacing[2]
+    if len(pixel_spacing) == 3:
+        image_geom.slice_thickness = pixel_spacing[2]
     image_geom.image_orientation = [1, 0, 0, 0, 1, 0]
 
     return image_geom
@@ -135,11 +147,19 @@ def e2e_dicom_metadata(
     meta.oct_image_params = e2e_image_params()
     if type(image) == OCTVolumeWithMetaData:
         meta.series_info = e2e_series_meta(
-            image.volume_id, image.laterality, image.acquisition_date
+            image.volume_id,
+            image.laterality,
+            image.acquisition_date,
+            image.metadata,
         )
         meta.image_geometry = e2e_image_geom(image.pixel_spacing)
     else:  # type(image) == FundusImageWithMetaData
-        meta.series_info = e2e_series_meta(image.image_id, image.laterality, None)
-        meta.image_geometry = e2e_image_geom([1, 1, 1])
+        meta.series_info = e2e_series_meta(
+            image.image_id,
+            image.laterality,
+            None,
+            image.metadata,
+        )
+        meta.image_geometry = e2e_image_geom(image.pixel_spacing)
 
     return meta

oct_converter/dicom/metadata.py

@@ -96,6 +96,9 @@ class SeriesMeta:
     acquisition_date: t.Optional[datetime.datetime] = None
     # Anatomy
     opt_anatomy: OPTAnatomyStructure = OPTAnatomyStructure.Unspecified
+    # Scan
+    protocol: str = ""
+    description: str = ""
 
 
 @dataclasses.dataclass

oct_converter/image_types/fundus.py

@@ -23,6 +23,8 @@ class FundusImageWithMetaData(object):
         patient_id: patient ID.
         image_id: image ID.
         DOB: patient date of birth.
+        metadata: all metadata parsed from the original file.
+        pixel_spacing: [x, y] pixel spacing in mm
     """
 
     def __init__(
@@ -33,13 +35,15 @@ def __init__(
         image_id: str | None = None,
         patient_dob: str | None = None,
         metadata: dict | None = None,
+        pixel_spacing: list[float] | None = None,
     ) -> None:
         self.image = image
         self.laterality = laterality
         self.patient_id = patient_id
         self.image_id = image_id
         self.DOB = patient_dob
         self.metadata = metadata
+        self.pixel_spacing = pixel_spacing
 
     def save(self, filepath: str | Path) -> None:
         """Saves fundus image.

oct_converter/readers/binary_structs/e2e_binary.py

@@ -1,13 +1,15 @@
 from construct import (
     Array,
     Float32l,
+    Float64l,
     Int8un,
     Int16un,
     Int32sn,
     Int32un,
     Int64un,
     PaddedString,
     Struct,
+    this,
 )
 
 # Mostly based on description of .e2e file format here:
@@ -77,7 +79,9 @@
     "patient_id" / PaddedString(25, "ascii"),
 )
 lat_structure = Struct(
-    "unknown" / Array(14, Int8un), "laterality" / Int8un, "unknown2" / Int8un
+    "unknown" / Array(14, Int8un),
+    "laterality" / PaddedString(1, "ascii"),
+    "unknown2" / Int8un,
 )
 contour_structure = Struct(
     "unknown0" / Int32un,
@@ -114,3 +118,98 @@
     "numAve" / Int32un,
     "imgQuality" / Float32l,
 )
+
+# Chunk 7: Eye Data (libE2E)
+eye_data = Struct(
+    "eyeSide" / PaddedString(1, "ascii"),
+    "iop_mmHg" / Float64l,
+    "refraction_dpt" / Float64l,
+    "c_curve_mm" / Float64l,
+    "vfieldMean" / Float64l,
+    "vfieldVar" / Float64l,
+    "cylinder_dpt" / Float64l,
+    "axis_deg" / Float64l,
+    "correctiveLens" / Int16un,
+    "pupilSize_mm" / Float64l,
+)
+
+# 9001 Device Name
+# Files examined have n_strings=3, string_size=256,
+# text=["Heidelberg Retina Angiograph", "HRA", ""]
+device_name = Struct(
+    "n_strings" / Int32un,
+    "string_size" / Int32un,
+    "text" / Array(this.n_strings, PaddedString(this.string_size, "u16")),
+)
+
+# 9005 Examined Structure
+# Files examined have n_strings=1, string_size=256,
+# text=["Retina"]
+examined_structure = Struct(
+    "n_strings" / Int32un,
+    "string_size" / Int32un,
+    "text" / Array(this.n_strings, PaddedString(this.string_size, "u16")),
+)
+
+# 9006 Scan Pattern
+# Files examined have n_strings=2, string_size=256,
+# and scan patterns including "OCT Art Volume", "Images", "OCT B-SCAN",
+# "3D Volume", "OCT Star Scan"
+scan_pattern = Struct(
+    "n_strings" / Int32un,
+    "string_size" / Int32un,
+    "text" / Array(this.n_strings, PaddedString(this.string_size, "u16")),
+)
+
+# 9007 Enface Modality
+# Files examined have n_strings=2, string_size=256,
+# and modalities including ["Infra-Red", "IR"],
+# ["Fluroescein Angiography", "FA"], ["ICG Angiography", "ICGA"]
+enface_modality = Struct(
+    "n_strings" / Int32un,
+    "string_size" / Int32un,
+    "text" / Array(this.n_strings, PaddedString(this.string_size, "u16")),
+)
+
+# 9008 OCT Modality
+# Files examined have n_strings=2, string_size=256, text=["OCT", "OCT"]
+oct_modality = Struct(
+    "n_strings" / Int32un,
+    "string_size" / Int32un,
+    "text" / Array(this.n_strings, PaddedString(this.string_size, "u16")),
+)
+
+# 10025 Localizer
+# From eyepy; "transform" is described as "Parameters of affine transformation"
+localizer = Struct(
+    "unknown" / Array(6, Float32l),
+    "windate" / Int32un,
+    "transform" / Array(6, Float32l),
+)
+
+# 3 seems to indicate the start of the chunk pattern
+# Examined files seem to have a mostly-regular pattern of 3, 2, ..., 5, 39
+# Both chunks 3 and 5 seem to include laterality info
+pre_data = Struct(
+    "unknown" / Int32un,
+    "laterality" / PaddedString(1, "ascii"),
+    # There's more here that I'm unsure of.
+    # There seems to be an "ART" in this chunk.
+)
+
+# 39 has some time zone data
+time_data = Struct(
+    "unknown" / Array(46, Int32un),
+    "timezone1" / PaddedString(66, "u16"),
+    "unknown2" / Array(9, Int16un),
+    "timezone2" / PaddedString(66, "u16"),
+    # There's more in this chunk (possibly datetimes, given tz)
+    # and the chunk size varies.
+)
+
+# 52, 54, 1000, 1001 seem to be UIDs with padded strings
+# 1000 may be StudyInstanceUID
+uid_data = Struct("uid" / PaddedString(64, "ascii"))
+
+# 1007 padded string with a brand name
+unknown_data = Struct("unknown" / PaddedString(64, "ascii"))

oct_converter/readers/boct.py

@@ -8,7 +8,7 @@
 
 import h5py
 import numpy as np
-from construct import Struct, StringError
+from construct import StringError, Struct
 from numpy.typing import NDArray
 
 from oct_converter.exceptions import InvalidOCTReaderError