DVH: Contours lying on dose-voxel edges produce “Empty DVH”

[SaharYaz]

When a structure’s contour lies exactly on a dose voxel edge, the DVH mask generation treats all voxel centres as outside the polygon, yielding an empty mask. For small/degenerate ROIs (e.g., single-voxel cases), _calculate_dvh then returns an “Empty DVH” (single zero bin) instead of a non-empty histogram, and the structure volume is not recorded in the final bin.

I've made a test that reproduces this issue.

def test_histogram_preserves_final_bin(self):
    """Regression test for maxdose computation."""
    # Create minimal RT Dose dataset with max dose slightly above an
    # integer bin boundary so rounding down would truncate it.
    dose = Dataset()
    dose.SOPClassUID = "1.2.840.10008.5.1.4.1.1.481.2"
    dose.SOPInstanceUID = generate_uid()
    dose.Modality = "RTDOSE"
    dose.Rows = 1
    dose.Columns = 1
    dose.NumberOfFrames = 1
    dose.GridFrameOffsetVector = [0.0]
    dose.FrameIncrementPointer = (0x3004, 0x000c)
    dose.ImagePositionPatient = [0.0, 0.0, 0.0]
    dose.ImageOrientationPatient = [1, 0, 0, 0, 1, 0]
    dose.PixelSpacing = [1.0, 1.0]
    dose.SamplesPerPixel = 1
    dose.PhotometricInterpretation = "MONOCHROME2"
    dose.BitsAllocated = 32
    dose.BitsStored = 32
    dose.HighBit = 31
    dose.PixelRepresentation = 0
    dose.DoseUnits = "GY"
    dose.DoseType = "PHYSICAL"
    dose.DoseSummationType = "PLAN"
    dose.DoseGridScaling = 0.00001

    arr = numpy.array([[[53001]]], dtype=numpy.uint32)
    dose.PixelData = arr.tobytes()

    file_meta = FileMetaDataset()
    file_meta.FileMetaInformationVersion = b"\x00\x01"
    file_meta.MediaStorageSOPClassUID = dose.SOPClassUID
    file_meta.MediaStorageSOPInstanceUID = dose.SOPInstanceUID
    file_meta.TransferSyntaxUID = "1.2.840.10008.1.2"
    dose.file_meta = file_meta
    dose.is_implicit_VR = True
    dose.is_little_endian = True

    # Define a simple square structure covering the single dose voxel
    structure = {
        "id": 1,
        "name": "square",
        "thickness": 1,
        "planes": {
            0.0: [
                {
                    "type": "CLOSED_PLANAR",
                    "num_points": 4,
                    "data": [
                        [0.0, 0.0, 0.0],
                        [1.0, 0.0, 0.0],
                        [1.0, 1.0, 0.0],
                        [0.0, 1.0, 0.0],
                    ],
                }
            ]
        },
    }

    dvh_data = dvhcalc._calculate_dvh(structure, dicomparser.DicomParser(dose))
    self.assertEqual(dvh_data.histogram.size, 54)
    # # Final bin should contain volume of single 1×1×1 mm voxel = 0.001 cm³
    expected_vol = (
        dose.PixelSpacing[0]
        * dose.PixelSpacing[1]
        * structure["thickness"]
    ) / 1000.0
    self.assertAlmostEqual(dvh_data.histogram[-1], expected_vol)

[bastula]

Thanks for finding this issue and creating a fix in the PR! Will have a look shortly and get back to you.