dicom_toolkit 0.2.7

An advanced medical imaging and DICOM processing library for Flutter.

0.2.7

• (fix) web: DicomToolkit.init() detects web platform and loads WASM from the Flutter asset path instead of the web root — no manual copy step needed in the consuming app
• (build) tool/rebuild_wasm.ps1 — convenience script for rebuilding WASM after Rust changes; handles RUSTFLAGS, wasm-pack, and .gitignore cleanup; release build produces ~1.3 MB WASM (down from 8.5 MB debug)
• (cleanup) removed example/web/pkg/ — no longer needed since WASM is bundled via package assets

0.2.6

• (fix) web: web/pkg/ and rust/pkg/ .gitignore files blocked WASM artifacts from git — consumers got empty directories on clone, causing wasm_bindgen is not defined + 404 errors at runtime
• (fix) web: dicom_toolkit.js + dicom_toolkit_bg.wasm are now committed to git — consumers get pre-built WASM automatically, no Rust toolchain required
• (fix) web: added web/pkg/ as a Flutter asset in pubspec.yaml — flutter build web now bundles WASM into assets/packages/dicom_toolkit/web/pkg/ automatically
• (docs) AGENTS.md: updated Build Commands, architecture diagram, Web WASM artifacts section, and version to 0.2.6
• (docs) README.md: updated Web support section with new rebuild script and asset bundling workflow

0.2.5

• (feat) 2 new extracted DICOM tags in Rust DicomMetadata (35→37): imageOrientationPatient (0020,0037), numberOfFrames (0028,0008)
• (feat) new DicomTagId constants (31→39): imageOrientationPatient, numberOfFrames, plus seriesDate, acquisitionDate, contentDate, toothNumber, toothRegion, imagerPixelSpacing (existing but untested)
• (feat) DicomMetadata typed getters: imageOrientationPatient, numberOfFrames
• (feat) DicomParseResult.frameCount now reads from metadata.numberOfFrames instead of hardcoded 1
• (feat) dual-path pixel extraction: PATH 1 reads raw PixelData element bytes directly (fast, no decoder needed for uncompressed DICOM), PATH 2 falls back to dicom-pixeldata decoder for compressed transfer syntaxes
• (feat) first-frame-only extraction: multi-frame datasets only process the first frame's pixel data, avoiding OOM and minutes-long decodes on large files
• (feat) convert_pixels() helper extracted in Rust — shared by both pixel paths
• (feat) DicomMetadata computed properties: isMultiFrame (numberOfFrames > 1), hasSpatialInfo (imagePositionPatient non-empty + slice spacing/thickness), isVolumetric (isMultiFrame || hasSpatialInfo)
• (feat) example app _MetadataTab: volumetric indicator tile, new Spatial section (pixel spacing, image position, slice location, slice spacing, orientation), and Frames tile in File section
• (fix) RangeError in DicomRenderer.createTexture when pixel buffer length exceeds width × height — loop now clamped to min(pixelCount, data.length)
• (fix) web: bytes.toList() memory explosion on large files — pass Uint8List directly to avoid 8× JS number unboxing
• (fix) web: FragmentProgram.fromAsset() unsupported on CanvasKit — shader compilation caught silently, falls back to CPU windowing via RawImage
• (fix) web: DicomViewer null shader crash — uses RawImage widget when GPU shader unavailable
• (fix) _MetadataTab missing group('Tooth info') header after refactor
• (build) WASM max memory: 1 GB → 2 GB to support large multi-frame DICOMs
• (perf) first-frame truncation avoids converting all frames to i16 and scanning all frames for window auto-range
• (debug) Rust eprintln! traces behind #[cfg(debug_assertions)] — auto-strip in release builds
• (debug) extensive Dart print() traces through decoder, renderer, controller, viewer, and example app
• (docs) DicomTagId count: 34→39; DicomMetadata getter count: 35→37; DicomParseResult tag count: 28→37
• (test) 259 tests pass; new computed property, spatial positioning, numberOfFrames, pixel spacing parsed component, and volumetric integration tests

0.2.4

• (feat) 3 new spatial positioning fields in Rust DicomMetadata (32→35): imagePositionPatient (0020,0032), sliceLocation (0020,1041), spacingBetweenSlices (0018,0088) — enables CBCT multi-slice volume reconstruction
• (feat) new DicomTagId constants (28→31): imagePositionPatient, sliceLocation, spacingBetweenSlices
• (feat) DicomMetadata typed getters: imagePositionPatient, imagePositionX/Y/Z, sliceLocation, spacingBetweenSlices
• (feat) convenience getters: pixelSpacingX, pixelSpacingY, instanceNumberInt
• (refactor) dicom_decoder.dart tag map updated with 3 new spatial tags
• (docs) README: 32→35 extracted tags, new Spatial category; AGENTS.md: version 0.2.0→0.2.3, tag/field counts updated
• (test) dicom_tag_id_test.dart: 28→31 constants; all existing test DicomMetadata constructors updated for new required fields

0.2.3

• (fix) JPEG-compressed DICOM decoding via dicom-pixeldata PixelDecoder trait — handles Carestream and other vendors using encapsulated transfer syntaxes (JPEG lossless, JPEG-LS, JPEG 2000, RLE)
• (fix) window center alignment for unsigned pixel data — DICOM window values are now offset by -32768 to match the shader's signed coordinate space, eliminating all-white/all-black images across Generic, Sirona, and Carestream files
• (fix) initial window defaults to pixel data range ("Full Range") instead of DICOM header values, which are often useless (e.g. C:32768/W:65536 spanning the full 16-bit range)
• (fix) window now resets on each loadFromBytes call; previously kept the previous file's window values
• (fix) level/width slider ranges computed from actual pixel data range instead of DICOM header values, eliminating ±40K dead zones
• (deps) added dicom-pixeldata = "0.9" to Rust dependencies for transfer-syntax-aware pixel decoding
• (build) WASM rebuilt with JPEG decoder support
• (test) 27 new tests: 23 integration tests against real Carestream/Generic/Sirona DICOM files validating JPEG decoding, window alignment, MONOCHROME1 detection, 12-bit-in-16-bit handling, and preset computation; 4 controller tests for auto-window and load-reset behavior

0.2.2

• (feat) expanded DICOM date extraction to 4 tags: studyDate (0008,0020), seriesDate (0008,0021), acquisitionDate (0008,0022), contentDate (0008,0023)
• (feat) DicomMetadata.bestDate — resolves the most recent valid non-future date among all four date tags; handles DICOM date ranges, leap years, and malformed values
• (feat) DicomMetadata.toothInfo — extracts tooth identification from dental DICOMs via cascade: standard Tooth Number (0018,6032), Tooth Region (0018,6033), then vendor private tags (Planmeca, Carestream, VATECH, Sirona, Dexis, KaVo), with generic fallbacks (ImageComments, AcquisitionProtocolName, AcquisitionContextDescription)
• (feat) new DicomTagId constants: seriesDate, acquisitionDate, contentDate, toothNumber, toothRegion
• (refactor) Rust DicomMetadata struct: 28→32 fields (added series_date, acquisition_date, content_date, tooth_info)
• (refactor) dicom_decoder.dart tag map updated with new date and tooth tag entries
• (docs) README: 28→32 extracted tags table, new Dates and Dental categories, bestDate usage section
• (docs) AGENTS.md: updated field count and tag counts throughout
• (test) 21 new tests: 11 bestDate edge cases, 5 tooth info tests, 5 date field delegation tests
• (example) metadata tab: new Dates section (best date + all 4 raw dates), Tooth tile in Acquisition section; viewer overlay uses bestDate

0.2.1

• Removed dart:io import completely. accepting only bytes as input.

0.2.0

• From here on, dicom_toolkit was born.
• (BREAKING) package renamed from flutter_dicom to dicom_toolkit
  • Update imports to package:dicom_toolkit/dicom_toolkit.dart
  • DicomController → DicomViewerController; DicomService removed
  • FileDicomLoader → RustDecoder (now implements DicomDecoder)
  • Loading is bytes-only on all platforms — loadFromBytes() only
• (BREAKING) composable toolkit API replaces monolithic controller
  • New: DicomParser, DicomRenderer, DicomRoi, DicomRuler, DicomExport, DicomWindowPreset
  • All tools accept dependency injection via abstract interfaces
• (feat) DicomTagId — 28 predefined DICOM tag constants + generic lookup via DicomMetadata.tag()
• (feat) DicomPixelData — sealed class hierarchy, DicomInt16PixelData for 16-bit monochrome
• (feat) DicomViewer widget — interactive flag to disable built-in gestures for measurement overlays
• (feat) rotation — 0°/90°/180°/270° via rotateClockwise()/rotateCounterClockwise(), included in PNG export
• (feat) pixel-spacing-aware DicomRuler — measures mm distance with fallback to Imager Pixel Spacing (0018,1164) for X-ray
• (feat) modality-agnostic DicomWindowPreset.forImage() — presets adapt to pixel value range, not hardcoded HU
• (feat) DicomExport.toPngBytes() — web-safe PNG export (no dart:io)
• (feat) DicomParser.parseMetadata() — metadata-only parse with skipPixels: true, no GPU needed
• (feat) web support — WASM build instructions, cross-origin isolation headers, serve_debug.py
• (feat) example app with measurement tools (ruler + ROI), rotation, presets, color maps, PNG export
• (fix) shader uniform index 8 bound for u_monochrome1 flag
• (fix) RangeError.range used in DicomParseResult.frame() instead of .index (compatibility)
• (refactor) 28-field Rust DicomMetadata struct (added pixel_spacing)
• (refactor) public_member_api_docs enforced — all public API documented
• (refactor) MedicalScreen removed — use DicomViewer directly
• (chore) analysis_options.yaml strict lint rules; cargokit excluded from analysis
• (docs) rewritten README.md and AGENTS.md reflecting new architecture
• (docs) 175 unit + widget + integration tests, all passing

0.1.0+3

• (docs): refine performance benchmarks in README

0.1.0+4

• (feat): add readDicomInfo() — lightweight metadata-only API (no controller, no shaders, no pixels)
• (feat): add DicomFileInfo class wrapping filePath, fileName, and DicomMetadata
• (feat): expand metadata extraction from 13 to 27 DICOM tags
  • Added patient demographics: studyDescription
  • Added equipment info: modality, manufacturer, manufacturerModelName, institutionName
  • Added identifiers: studyInstanceUid, seriesInstanceUid, sopInstanceUid
  • Added acquisition details: seriesDescription, bodyPartExamined, sliceThickness, instanceNumber
• (refactor): introduce get_str_tag, get_float_tag, get_int_tag helpers for cleaner tag extraction
• (feat): add 8-bit pixel data support with signedness-aware conversion
• (perf): reorder pixel extraction to try fast byte-chunking first, with transfer-syntax-aware fallback
• (docs): update README with 27-field metadata table, readDicomInfo() guide, and usage examples

0.1.0+3

• (docs): refine performance benchmarks in README

0.1.0+2

• (feat): add comprehensive performance benchmarks and metadata defaults
• (feat) add DicomMetadata defaults and refactor DICOM processing
• (feat): implement DicomMetadata defaults and a new constructor in Rust; refactor process_dicom_file to use them.
• Added Default implementation and a new constructor for DicomMetadata in Rust.
• Refactored process_dicom_file to use DicomMetadata defaults when tags are missing.
• (feat): add newInstance and default_ static methods to DicomMetadata and DicomFrameResult Dart classes.
• Added newInstance and default_ static methods to DicomMetadata and DicomFrameResult Dart classes.
• (feat): introduce LibShaders constant for centralized shader asset path management.
• Introduced LibShaders constant for centralized shader asset path management.
• (refactor): integrate LibShaders into DicomController.
• Integrated LibShaders into DicomController for shader loading.
• (fix): update MedicalScreen to support an optional title parameter.
• Updated MedicalScreen to support an optional title parameter.
• (feat): add series_performance_test.dart covering throughput, latency, windowing stress, and lifecycle tests.
• (docs): add detailed performance benchmarks and workstation-grade metrics to README.
• (chore): add test series data to .gitignore.

0.1.0+1

• (fix): README images src

0.1.0

• Initial release of Flutter-Dicom.
• High-performance Rust core for DICOM parsing and pixel extraction.
• GPU Fragment Shaders for real-time Windowing (Level/Width).
• Support for 16-bit precision and Hounsfield Unit (HU) mapping.
• Built-in DicomViewer widget with interactive pan, zoom, and contrast adjustments.
• Comprehensive metadata extraction (Patient Name, Modality, Rescale Slope/Intercept, etc.).
• Cross-platform support (Android, iOS, macOS, Windows, Linux).