Parse a Dicom formatted datetime string and return a Python datetime.

parse_dicom_dt.py

import datetime as dt
import pytest
import re


def parse_dicom_dt(dicom_dt):
    """
    Parse a Dicom formatted datetime string and return a Python datetime.

    The Dicom format is "YYYYMMDDHHMMSS.FFFFFF&ZZXX" described here:
    https://dicom.nema.org/medical/dicom/current/output/chtml/part05/sect_6.2.html#table_6.2-1
    """
    pattern = (
        "(\d{14})"                    # YYYYMMDDHHMMSS
        "(\.\d+)?"                    # .FFFFFF 6 digits of fractional seconds (optional)
        "(?:([+-])(\d{2})(\d{2})?)?"  # UTC offset (optional)
        "$"                           # end of string
    )

    m = re.match(pattern, dicom_dt)
    if m:
        try:
            # YYYYMMDDHHMMSS
            d1 = dt.datetime.strptime(m.group(1), "%Y%m%d%H%M%S")

            fractional_seconds = dt.timedelta(
                seconds=float("0"+m.group(2)) if m.group(2) else 0
            )

            # UTC offset
            offset_sign = -1 if m.group(3) == "-" else 1
            offset_hrs = offset_sign * int(m.group(4) or "0")
            offset_mins = offset_sign * int(m.group(5) or "0")
            offset = dt.timedelta(hours=offset_hrs, minutes=offset_mins)

            # Current advice is to use tz aware datetimes.
            # See: https://docs.python.org/3/library/datetime.html
            return (d1 + fractional_seconds - offset).replace(tzinfo=dt.timezone.utc)
        except ValueError as e:
            raise ValueError(f"Can't parse \"{dicom_dt}\" as a date.") from e
    else:
        raise ValueError(f"Can't parse \"{dicom_dt}\" as a date.")

EXAMPLES = {
    "20220102112233.123456-0800": dt.datetime(2022, 1, 2, 11, 22, 33, 123456, tzinfo=dt.timezone.utc) + dt.timedelta(hours=8),
    "20220102112233.01+1200": dt.datetime(2022, 1, 2, 11, 22, 33, 10000, tzinfo=dt.timezone.utc) + dt.timedelta(hours=-12),
    "20220102112233+1200": dt.datetime(2022, 1, 2, 11, 22, 33, tzinfo=dt.timezone.utc) + dt.timedelta(hours=-12),
    "20220102112233.1": dt.datetime(2022, 1, 2, 11, 22, 33, 100000, tzinfo=dt.timezone.utc),
    "20220102112233": dt.datetime(2022, 1, 2, 11, 22, 33, tzinfo=dt.timezone.utc),
}

BAD_EXAMPLES = [
    "",
    "Boink!",
    "20221399452233.123456-0800",
    "202201021122Z3.123456-0800",
    "20220102112233..123456-0800",
    "20220102112233.123456@0800",
    "20220102112233.123456-800",
    "20220102112233.123456-12345",
]

def test_parse_dicom_dt():
    for example, expected in EXAMPLES.items():
        x = parse_dicom_dt(example)
        assert x == expected, f"{x} not equal to {expected}"

def test_bad_dicom_dts():
    for example in BAD_EXAMPLES:
        with pytest.raises(Exception) as exc_info:
            parse_dicom_dt(example)

        print(exc_info.value)

if __name__ == "__main__":
    test_parse_dicom_dt()
    test_bad_dicom_dts()

This is a silly bit of code I wrote that is just slightly more flexible than using the strptime and strftime functions from the datetime standard library package along with the following format string:

DICOM_DATETIME_FORMAT = "%Y%m%d%H%M%S.%f%z"

I'm keeping this code around in case I find corner-cases not handled by that.

For details on Dicom's DateTime value representation "DT" see:
https://dicom.nema.org/medical/dicom/current/output/chtml/part05/sect_6.2.html#table_6.2-1