Dicomdiff

A comprehensive Python module for analyzing DICOM file pseudonymization and de-identification processes. Dicomdiff enables researchers and healthcare professionals to evaluate, compare, and validate different pseudonymization methods by providing detailed analysis of how DICOM tags are transformed during de-identification.

What Dicomdiff Does

Core Functionality:

• Individual File Comparison: Compare original DICOM files with their de-identified versions to detect specific changes
• Pseudonymization Method Analysis: Evaluate and compare different pseudonymization approaches across multiple files
• Cross-Method Comparison: Analyze how different pseudonymization tools handle the same source data
• Consistency Validation: Check whether pseudonymization methods behave consistently across datasets
• Tag Transformation Tracking: Monitor how specific DICOM tags are modified, removed, or preserved during pseudonymization

Installation

Install the module using pip

  pip install dicomdiff

Usage

Compare two DICOM files in Python

from dicomdiff.main import compare_dicom_files, print_differences

original_file = "path to original dcm file"
deidentified_file = "path to de-identified dcm file"

result = compare_dicom_files(original_file, deidentified_file) # compare the files
print_differences(result) # print the results

Compare two pseudonymization methods

import os
import glob
import pandas as pd

# Import the pseudonymizers you'll be comparing
from dicomdiff.pseudonymizer import IDISPseudonymizer, DicomRakePseudonymizer, InferredPseudonymizer
from dicomdiff.summary import generate_pseudonymization_summary

# Define the paths
input_dir = "path/to/input/data"
mapping_csv = "path/to/mapping"
idisoutput_dir = "/path/to/output/data/" # Note: this is needed if you use InferredPseudonymizer

# Find all files 
dicom_files = glob.glob(os.path.join(input_dir, "**", "*.dcm"), recursive=True)

# Define the pseudonymizers you want to use
dicomrake_pseudonymizer = DicomRakePseudonymizer()
idis_pseudonymizer = InferredPseudonymizer.from_csv(mapping_csv, idisoutput_dir)

# Generate comparison summary between the two pseudonymization methods
summary_df = generate_pseudonymization_summary(
    file_paths=dicom_files,
    pseudonymizer_a=dicomrake_pseudonymizer,
    pseudonymizer_b=idis_pseudonymizer,
    pseudonymizer_a_name="DICOMRake", # If you want to change the name of the pseudonymizers
    pseudonymizer_b_name="IDISPseudonymizer",
)

# Define helper function to identify private DICOM tags (odd group numbers)
def is_private_tag(tag_str):
    try:
        group = int(tag_str.split(",")[0], 16)
        return group % 2 != 0  
    except (ValueError, IndexError):
        return False

# Filter results to show only public tags (even group numbers)
public_tags_df = summary_df[~summary_df["tag"].apply(is_private_tag)]

# Display overall comparison statistics for public tags
print("\nComparison results (public tags only):")
print(public_tags_df["comparison"].value_counts().to_string())

# Configure pandas display options for better output formatting
pd.set_option("display.max_rows", None)
pd.set_option("display.max_columns", None)
pd.set_option("display.width", 1000)
pd.set_option("display.expand_frame_repr", False)
pd.set_option("display.colheader_justify", "center")

# Show detailed differences in public tags where pseudonymizers behaved differently
different_public_tags = public_tags_df[public_tags_df["comparison"] != "Both Unchanged"]
print(f"\nDifferences in public tags ({len(different_public_tags)}):")
print(different_public_tags)

# OPTIONAL: If private tags exist, you can show differences in private tags as well
if len(summary_df) > len(public_tags_df):
    private_tags_df = summary_df[summary_df["tag"].apply(is_private_tag)]
    different_private_tags = private_tags_df[
        private_tags_df["comparison"] != "Both Unchanged"
    ]
    print(f"\nDifferences in private tags ({len(different_private_tags)}):")
    print(different_private_tags)

Compare two DICOM files using CLI

# Compare two DICOM files
dicomdiff compare file1.dcm file2.dcm

# Filter results
dicomdiff compare file1.dcm file.dcm --changed

CLI Flags