Comprehensive 'DICOM' Toolkit for Medical Imaging

A comprehensive R interface to the 'DICOM' Toolkit ('DCMTK'), providing native R bindings for reading, writing, modifying, and visualizing Digital Imaging and Communications in Medicine ('DICOM') files. Features include 'tidyverse'-compliant data structures, 'ggplot2'-based image visualization, interactive 'Shiny' viewers, 'PACS' networking (C-ECHO, C-FIND, C-STORE, C-MOVE), structured report handling, 'DICOM' deidentification, and integration with local and cloud-based large language models ('LLM') via 'Ollama' and 'OpenAI'-compatible APIs for AI-assisted medical image analysis. The 'DCMTK' C++ library is bundled and compiled from source, requiring no external installation.

dicomtk: Comprehensive 'DICOM' Toolkit for Medical Imaging in R

dicomtk is a complete, modern rebuild of the R dcmtk wrapper, synchronized with OFFIS DCMTK 3.7.1 (PS 3.6-2026d) and powered by a bundled C++17 / Rcpp engine requiring zero external installation.

Installation

Install pre-compiled binaries or source directly from R-Universe:

install.packages(
  "dicomtk",
  repos = c(
    "https://lauyeehow1986-hub.r-universe.dev",
    "https://cloud.r-project.org"
  )
)

Or install the development version from GitHub:

# install.packages("remotes")
remotes::install_github("lauyeehow1986-hub/dicomtk")

Key Features

Zero External Dependencies: Bundles the full 5,354-tag DCMTK 3.7.1 data dictionary, headers (inst/include/dcmtk/), and native C++17 DICOM engine directly inside the package.
Tidyverse Compliant: All tabular functions (read_dicom(), dicom_tags(), dicom_find(), read_dicom_sr(), dicom_report()) return tibble objects (tbl_df) and compose naturally with |> and dplyr/tidyr.
Full DCMTK Module Coverage:
- Core (dcmdata): read_dicom(), read_dicom_header(), write_dicom(), dicom_tags(), dicom_modify(), dicom_tag_value(), dicom_tag_lookup(), dicom_anonymize(), dcm2xml(), dcm2json(), dcm2pdf().
- Image Processing & Codecs (dcmimgle, dcmimage, dcmjpeg, dcmjpls): read_dicom_pixels(), read_dicom_series(), dicom_window(), dicom_compress(), dicom_decompress(), dicom_to_png(), dicom_to_nifti().
- Visualization: plot_dicom() and ggdicom() (ggplot2 with clinical CT/MR window presets), dicom_viewer() (interactive Shiny app), and render_volume() (3-plane MPR, MIP, and interactive rgl 3D rendering).
- PACS Networking (dcmnet, dcmqrdb, dcmwlm): dicom_connect(), dicom_echo(), dicom_find(), dicom_get(), dicom_send(), dicom_receive().
- Structured Reporting & Specialized IODs (dcmsr, dcmrt, dcmseg, dcmpmap, dcmtract, dcmect, dcmsign): read_dicom_sr(), create_dicom_sr(), read_dicom_rt(), read_dicom_seg(), read_dicom_pmap(), read_dicom_ect(), dicom_verify_signature().
Local & Cloud LLM Integration:
- Native support for local Ollama (http://localhost:11434), OpenAI (gpt-4o, etc.), Anthropic Claude (claude-sonnet-4-20250514, etc.), and any OpenAI-compatible endpoint (vLLM, LM Studio, OpenRouter, Azure).
- Automatic HIPAA Safe Harbor de-identification before sending metadata or images to remote APIs.
100% Backward Compatible: Preserves legacy dcmtk functions (dcmdump(), dcmhd(), wide_hdr(), burned_annot(), dcmodify(), dcm_anon(), dcm_subtract_date(), dcmj2pnm(), dcmdjpeg(), dcmdrle(), dcmcrle()).

Quick Start

library(dicomtk)
library(dplyr)

dcm_file <- system.file("extdata", "example.dcm", package = "dicomtk")

# 1. Read DICOM header as a tidy tibble
read_dicom(dcm_file) |>
  filter(group == "0028") |>
  select(tag, vr, keyword, value)

# 2. Visualize with ggplot2 and clinical CT/MR window presets
plot_dicom(dcm_file, window = "auto", palette = "gray")

# 3. Launch interactive Shiny DICOM viewer
if (interactive()) {
  dicom_viewer(dcm_file)
}

# 4. Connect to Local Ollama or OpenAI / Claude for AI-Assisted Analysis
llm <- dicom_llm("ollama", model = "llava:13b")
dicom_summarize(dcm_file, llm = llm)
dicom_describe_image(dcm_file, llm = llm)
dicom_report(dcm_file, llm = llm)