site address:
github.com/dmandache/IMPERANDI redirected to: github.com/dmandache/IMPERANDI
site title:
GitHub - dmandache/IMPERANDI: Python CLI that turns heterogeneous DICOM archives into analysis-ready CT/MR cohorts: NIfTI conversion, TotalSegmentator segmentation, contrast-phase curation, registration, PyRadiomics, QC viewer · GitHub
|
|
|
Our opinion (on Tuesday 17 November 2026 10:30:38 UTC):
- no comments
|
|
|
|
After content analysis of this website we propose the following hashtags:
|
|
|
|
Meta tags:
description= Python CLI that turns heterogeneous DICOM archives into analysis-ready CT/MR cohorts: NIfTI conversion, TotalSegmentator segmentation, contrast-phase curation, registration, PyRadiomics, QC viewer - dmandache/IMPERANDI;
Headings (most frequently used words):
and, navigation, repository, contributing, about, files, footer, topics, resources, stars, watchers, forks, dicom, menu, latest, commit, history, folders, google, colab, demos, start, here, what, imperandi, does, installation, quickstart, configuration, outputs, cohort, selection, resuming, retries, logging, quality, control, troubleshooting, layout, further, examples, development, license, disclaimers, support, releases, packages, used, by, contributors, languages, optional, features, install, from, source, inspect, ingest, metadata, convert, volumes, to, nifti, segment, images, resolve, contrast, phases, optionally, register, organs, masks, extract, radiomics, research, use, only, privacy, funding,
Text of the page (most frequently used words):
and (139), the (91), csv (49), for (47), imperandi (46), phase (34), project (29), dicom (26), with (26), manifest (25), tables (23), #install (22), cohort (21), radiomics (20), github (18), selection (18), viewer (17), python (17), curation (17), registration (17), use (17), pip (17), nifti (16), your (16), mask (16), you (15), segmentation (15), source (15), all (15), from (14), images (13), are (13), generic (13), run (13), pyradiomics (12), optional (12), ingest (12), path (12), metadata (12), not (11), repository (11), contributing (11), data (11), input (11), output (11), table (11), workflow (10), command (10), settings (10), paths (10), volume (10), files (10), when (10), segment (10), analysis (9), conversion (9), columns (9), image (9), convert (9), csv_path_out (9), yaml (9), organ (9), features (9), masks (9), can (8), that (8), this (8), docs (8), totalsegmentator (8), before (8), processing (8), see (8), tests (8), dataset (8), git (8), after (8), stage (8), license (7), liver (7), cleaning (7), into (7), contrast (7), support (7), only (7), reports (7), examples (7), rules (7), build (7), guide (7), pipeline (7), volumes (7), identifiers (7), output_dir (7), register (7), built (7), installation (7), mri (6), imaging (6), cli (6), archives (6), work (6), outputs (6), documentation (6), their (6), public (6), start (6), filters (6), rows (6), resume (6), nifti_path (6), level (6), csv_path (6), parse (6), provenance (6), per (6), canonical (6), one (6), tasks (6), code (6), security (5), readme (5), its (5), reference (5), check (5), dev (5), environment (5), also (5), full (5), src (5), these (5), more (5), phases (5), original (5), inspect (5), steps (5), nifti_index_phased (5), installed (5), quality (5), root_path (5), configuration (5), another (5), quickstart (5), step (5), main (5), review (5), input_stem (5), modality (5), baseline (5), tumor (5), stages (5), navigation (4), report (4), forks (4), stars (4), apache (4), resources (4), preprocessing (4), readthedocs (4), heterogeneous (4), ready (4), cohorts (4), research (4), without (4), development (4), notebook (4), html (4), keep (4), runs (4), dependencies (4), include (4), powershell (4), includes (4), interactive (4), dashboards (4), hooks (4), dataset_configs (4), manifests (4), error (4), default (4), while (4), retry_failed (4), including (4), extra (4), options (4), explore (4), tag (4), import (4), mask_ (4), _path (4), file (4), change (4), same (4), dicom_index_clean (4), nifti_index (4), checkpoint (4), uses (4), cleaned (4), fallback (4), model (4), each (4), clean (4), specific (4), consensus (4), then (4), extract (4), venv (4), demo (4), signed (4), copilot (4), status (3), medical (3), scans (3), clinical (3), turns (3), about (3), star (3), under (3), operandi (3), validation (3), open (3), index (3), running (3), slow (3), https (3), com (3), checkout (3), workflows (3), example (3), tcga (3), lihc (3), blueprint (3), colab (3), package (3), memory (3), space (3), conv_errors (3), archive (3), id_extraction (3), missing (3), first (3), must (3), does (3), browser (3), controls (3), discovers (3), nifti_index_registered (3), scanviewer (3), overlays (3), log (3), content (3), retains (3), recorded (3), failures (3), successful (3), available (3), names (3), produces (3), exports (3), separate (3), want (3), process (3), unresolved (3), representative (3), writes (3), individual (3), configured (3), patient_key (3), precedence (3), acquisition (3), alignment (3), groups (3), site (3), dry (3), copy (3), computed (3), point (3), ingestion (3), based (3), add (3), dmandache (3), test (3), feature (3), activate (3), folders (3), series (3), what (3), commit (3), actions (3), tab (3), window (3), refresh (3), session (3), reload (3), sign (3), developers (3), changelog (3), customer (3), enterprise (3), grade (3), community (2), privacy (2), footer (2), watching (2), watchers (2), activity (2), cite (2), radiology (2), pancreas (2), abdominal (2), topics (2), increase (2), supported (2), anr (2), plan (2), intended (2), conventions (2), pull (2), requests (2), reproducible (2), requirements (2), lint (2), fast (2), backed (2), follow (2), aim (2), harvard (2), master (2), virtual (2), activated (2), bash (2), slurm (2), entry (2), benchmarks (2), isolated (2), notebooks (2), demos (2), builtin_datasets_config (2), extraction (2), shared (2), contents (2), cases (2), detail (2), version (2), relevant (2), troubleshooting (2), reduce (2), num_workers (2), has (2), repeat (2), failed (2), disk (2), members (2), parsed (2), length (2), snapshot_tags (2), matching (2), display (2), scan (2), col (2), registered (2), instead (2), jupyter (2), patients (2), control (2), behavior (2), help (2), debug (2), logging (2), every (2), needed (2), checkpoints (2), strict_resume (2), preserving (2), detection (2), flush (2), long (2), derive (2), even (2), location (2), subsequent (2), directories (2), retain (2), dicom_index_curated (2), next (2), evidence (2), inspection (2), downstream (2), _curated (2), three (2), row (2), group (2), register_qc (2), unless (2), supplied (2), dicom_index (2), results (2), may (2), instances (2), dicom_path (2), operations (2), parameters (2), take (2), over (2), column (2), focused (2), range (2), order (2), reconstructed (2), grouping (2), post (2), strategy (2), mappings (2), ordered (2), filtering (2), tags (2), across (2), accept (2), utf (2), encoding (2), own (2), entire (2), patient (2), study (2), native (2), common (2), optionally (2), organs (2), result (2), stop (2), prediction (2), such (2), nifti_index_segmented (2), resolve (2), existing (2), within (2), place (2), crop (2), worker (2), small (2), plus (2), commands (2), containing (2), base (2), indexes (2), extras (2), using (2), requires (2), separately (2), runtime (2), viewers (2), needs (2), continuation (2), line (2), resolution (2), customize (2), adapt (2), try (2), light (2), items (2), pyproject (2), toml (2), citation (2), cff (2), gitignore (2), gitattributes (2), static (2), githooks (2), last (2), message (2), name (2), 620 (2), commits (2), latest (2), menu (2), insights (2), projects (2), universe (2), training (2), integrations (2), stories (2), center (2), learn (2), business (2), fixes (2), products (2), secret (2), perform, action, time, share, personal, information, manage, cookies, contact, terms, 2026, inc, languages, contributors, used, packages, releases, our, giving, repo, visibility, performed, rhu, was, part, french, national, agency, agence, nationale, recherche, 3rd, pia, integrated, france, 2030, rhus, 0012, funding, should, assumed, identify, users, responsible, appropriate, anonymization, governance, procedures, sharing, software, device, diagnosis, treatment, decisions, independent, disclaimers, distributed, cleanup, bug, contributions, welcome, txt, sphinx, going, coverage, minimum, require, prepared, them, setup, ruff, pytest, benchmark, runners, points, job, ircad, unit, user, guides, api, exploratory, local, shipped, modules, utilities, layout, further, covers, issue, described, method, diagnosing, spawn, workers, hang, exhausted, arterial, portal, venous, fewer, resolving, errors, skipped, indexed, cache, permissions, fails, orientation, spacing, thickness, removes, expected, depth, sample, unexpected, dependency, checks, symptom, accessible, server, uploading, upload, referenced, pypi, installations, snapshots, statistics, panel, serve, viewer_web, show, args, constructing, displays, automatically, supports, both, load, phase_col, scan_col, read_csv, pandas, widget, matplotlib, readable, values, provide, exams, reviewing, slices, adjustable, global, subcommand, changes, invalidate, cached, fingerprints, detect, verification, normal, fixing, cause, rerun, retry, hashing, stronger, ignore, state, fresh, no_resume, seconds, checkpoint_every_sec, processed, checkpoint_every_rows, option, compatible, interruption, resuming, retries, filenames, therefore, offers, override, replace, selected_csv_path, nifti_index_segmented_curated, phased, candidates, select, required, rule, qc_unresolved_phase, wide, scores, rather, than, qc_, contains, best, eligible, sequence, slot, exam, additional, gives, distinct, counts, finish, successfully, fail, radiomics_errors, _radiomics, logs, register_errors, _registered, phase_errors, warnings, applicable, seg_errors, _clean, dicom_tags_snapshot, ndjson, dicom_index_errors, preserves, accumulates, artifact, core, represent, multiple, assume, ordinary, volume_id, series_id, study_id, explains, ranking, loaded, explicit, pyradiomics_settings, differ, resegmentation, omits, execute, resolves, geometry, 1700, protocols, derives, patientid, from_tag, patientname, choices, adapting, new, modalities, phase_curation, ordering, derived_columns, id_standardization, identifier, auto, section, edit, consistently, read_text, write_text, joinpath, pathlib, importlib, pass, bundled, json, accepted, annotated, shape, intensity, texture, paired, subtracted, empty, ran, populated, applies, nifti_index_radiomics, publishes, grid, remain, fields, segmentations, source_, output_space, moving, during, need, installing, backend, tries, predictor, refreshed, phase_reason, phase_source, crops, foreground, bounding, box, voxels, bounds, cropping, disabled, margin, lesion, counterparts, producing, starting, limits, simultaneous, validating, mask_liver_tumor_path, mask_liver_path, receive, conversions, reported, processes, retained, export, sampled, snapshot, resolved, directory, collection, converted, artifacts, dicoms, generated, editable, installs, clone, download, weights, allow, network, access, storage, those, downloads, size, details, downloader, formatting, tools, packaged, simpleitk, windows, multiline, shell, below, syntax, enter, backtick, equivalent, scripts, ps1, bin, upgrade, create, slice, selected, category, browse, roi, prefixed, align, combine, labels, through, ontology, predictions, logical, morphological, curated, dicom2nifti, references, followed, normalize, dates, filter, unsuitable, curate, sequences, instance, discover, zip, tar, headers, purpose, root, folder, choose, matches, here, adds, torchio, pytorch, batch, launch, google, reads, linked, provides, framework, interface, turning, collections, traceable, standardizes, curates, connects, normalization, diagnostic, interoperability, view, date, history, branches, fork, notification, notifications, discussions, issues, dismiss, alert, switched, accounts, out, appearance, search, pricing, event, talks, workshops, partners, company, forum, events, webinars, blog, resource, connect, premium, advanced, ons, trust, why, evaluate, industries, government, manufacturing, financial, services, healthcare, industry, nonprofits, startups, team, enterprises, businesses, improvements, developer, articles, sponsors, trending, repos, certifications, scanning, leak, prevention, protection, vulnerability, remediation, secure, cloud, environments, codespaces, enforcement, automation, ship, coding, agent, automated, hosting, collaboration, platform, skip,
Text of the page (random words):
path that matches what you want to do explore the workflow try the light colab demo on public data process your own cohort follow the quickstart point root_path to the root folder containing your dicom files adapt imperandi to your dataset copy the manifest blueprint and customize it for your identifiers and curation rules you can start with metadata curation then add conversion segmentation registration or radiomics extraction as needed what imperandi does stage purpose main result parse discover dicom folders and zip tar archives extract headers and derive patient study series identifiers instance level dicom index clean group instances into volumes normalize dates and acquisition order filter unsuitable images and curate ct mr sequences and phases volume level cohort and optional representative series selection ingest run parse followed by clean parsed and cleaned indexes in one workflow convert convert curated volumes with dicom2nifti while preserving source references nifti images and nifti_path columns segment run modality specific totalsegmentator tasks and optional logical morphological mask operations mask_ output _path columns and nifti masks phase resolve canonical contrast phase through ordered rules ontology mappings and optional model predictions phase labels provenance selection and qc tables register align organs within configured groups and optionally combine masks in native or common reference space registered consensus masks original paths and registration qc radiomics extract pyradiomics features using modality specific settings and cohort filters roi prefixed feature columns qc viewers browse images and mask overlays in jupyter or a browser interactive review of the cohort dicom folders archives parse dicom index slice level metadata csv clean selected cohort volume level csv phase phase provenance selection per category convert nifti images segment organ tumor masks register organ alignment tumor consensus radiomics feature table phase resolution can also run before segmentation radiomics needs image and mask paths plus a phase column when phase based filtering is configured long running stages support checkpoints and resume with per volume failures recorded separately see the workflow guide for stage dependencies and cohort selection installation requires python 3 10 3 13 create an isolated environment and install the base package for dicom ingest and nifti conversion python m venv venv source venv bin activate pip install upgrade pip pip install imperandi imperandi help on windows powershell activate with venv scripts activate ps1 the multiline shell examples below use bash continuation syntax in powershell enter each command on one line or use powershell s backtick continuation python m imperandi is equivalent to the imperandi entry point optional features install the features your workflow needs into the same environment feature install command segmentation and image based phase prediction pip install imperandi segment organ registration with simpleitk pip install imperandi registration notebook and browser qc viewers pip install imperandi viewer all packaged runtime extras including registration dependencies pip install imperandi all test lint and formatting tools pip install imperandi dev public dataset test downloader pip install imperandi slow runtime development and dataset test dependencies pip install imperandi all dev pyradiomics is installed separately including when using all or all dev the project s installation workflow uses git and requires git to be available pip install pyradiomics git https github com aim harvard pyradiomics git master totalsegmentator may download model weights on first use allow network access and storage for those downloads and start with a small cohort to size memory and worker requirements installation details are in the installation guide install from source with your virtual environment activated git clone https github com dmandache imperandi git cd imperandi pip install e for editable installs with extras use commands such as pip install e segment viewer or pip install e all dev quickstart keep source dicoms cohort tables and generated images in separate directories project dicom source dicom collection tables cohort indexes error reports and qc tables nifti converted images and segmentation outputs registration optional registration artifacts run these examples from the directory containing project use the base installation for steps 1 2 the segment extra for steps 3 4 the registration extra for step 5 and the separate pyradiomics installation for step 6 1 inspect and ingest dicom metadata all pipeline commands accept dry run to inspect their plan and resolved settings before processing imperandi ingest root_path project dicom output_dir project tables manifest generic dry run then run ingestion imperandi ingest root_path project dicom output_dir project tables manifest generic this writes dicom_index csv and dicom_index_clean csv plus selection and qc reports the cleaned table has one row per retained reconstructed volume add snapshot_tags to export a sampled dicom tag snapshot for metadata inspection 2 convert volumes to nifti imperandi convert csv_path project tables dicom_index_clean csv output_dir project nifti csv_path_out project tables nifti_index csv successful rows receive nifti_path failed conversions are reported in conv_errors csv this example processes the full cleaned cohort to process representative volumes instead use dicom_index_curated csv as input see cohort selection 3 segment images imperandi segment csv_path project tables nifti_index csv csv_path_out project tables nifti_index_segmented csv manifest generic num_workers 1 the built in manifest runs liver and liver lesion tasks for ct and their mr counterparts for mr mri producing mask_liver_path and mask_liver_tumor_path starting with one worker limits simultaneous model and image memory use increase it after validating a small run optional crop crop margin mm 50 crops the source nifti and all output masks in place to a shared foreground bounding box before mask post processing it retains only existing voxels within the image bounds cropping is disabled in the built in manifest 4 resolve contrast phases imperandi phase csv_path project tables nifti_index_segmented csv csv_path_out project tables nifti_index_phased csv manifest generic the built in strategy tries metadata rules first then totalsegmentator for unresolved ct rows mr phase curation uses metadata the built in image based predictor is ct only outputs include canonical phase provenance such as phase_source and phase_reason and refreshed selection qc tables metadata curation also runs during generic ingestion if you only need that result you can stop after ingest without installing a prediction backend 5 optionally register organs and masks imperandi register csv_path project tables nifti_index_phased csv csv_path_out project tables nifti_index_registered csv output_dir project registration manifest generic the built in manifest groups scans by patient study and modality and publishes masks on each scan s native grid output_space moving for successful rows canonical mask paths point to the consensus results original paths remain in source_ columns review register_qc csv and status fields before downstream analysis see registration for common reference output alignment stages and options to retain individual segmentations 6 extract radiomics for the workflow without registration imperandi radiomics csv_path project tables nifti_index_phased csv csv_path_out project tables nifti_index_radiomics csv manifest generic if you ran step 5 use nifti_index_registered csv as the input to extract from its masks radiomics discovers populated mask_ _path columns and applies the manifest s ct or mr settings organ shape features are computed from the entire organ mask intensity and texture features are computed from the organ mask with its paired tumor mask subtracted when available if the tumor mask is missing or empty all features are computed from the entire organ mask configuration pass manifest generic for the bundled baseline or manifest site a yaml for your own yaml file json manifests are not accepted start from the built in manifest or the annotated blueprint to copy the installed baseline without a source checkout run this in python from importlib resources import files from pathlib import path baseline files imperandi builtin_datasets_config joinpath manifests generic yaml path site a yaml write_text baseline read_text encoding utf 8 encoding utf 8 edit the copy then use it consistently across the stages that accept a manifest imperandi ingest root_path project dicom output_dir project tables manifest site a yaml dry run manifest section controls id_extraction identifier source auto tags or path dicom tag names and path fallback id_standardization derived_columns python hooks for dataset specific identifiers and metadata cleaning steps ordered filtering volume grouping acquisition ordering modality curation and phase curation phase_curation strategy precedence canonical phase mappings unresolved fallback and selection groups segmentation modalities ct mr tasks output names and optional mask post processing registration grouping organ tumor mask columns alignment stages consensus and output space radiomics ct mr pyradiomics parameters and cohort filters review these baseline choices when adapting to a new dataset generic derives patient_key from patientname with path fallback change id_extraction patient_key from_tag if your cohort uses patientid or another tag cleaning includes geometry and acquisition filters including a reconstructed volume length range of 30 1700 mm inspect these rules for your protocols cleaning steps execute in order modality curation produces evidence the subsequent phase curation step resolves canonical phases and selection segmentation tasks are liver focused and radiomics settings differ for ct and mr the ct configuration includes an hu resegmentation range mr omits it manifest radiomics parameters take precedence over pyradiomics_settings and manifest filters take precedence over cli filters for the same column the configuration guide explains hooks phase rules selection ranking mask operations and validation repository specific operandi settings in dataset_configs are loaded by explicit yaml path and are not installed with the package outputs and cohort selection the main cohort table preserves source identifiers and accumulates artifact paths and results core columns include patient_key study_id series_id volume_id dicom_path nifti_path mask_ output _path and canonical phase provenance dicom_path may represent multiple instances or archive members do not assume it is one ordinary file path stage default main table error or qc reports parse dicom_index csv dicom_index_errors csv optional dicom_tags_snapshot ndjson clean ingest input_stem _clean csv dicom_index_clean csv selection and phase qc exports when configured convert nifti_index csv conv_errors csv segment input csv unless csv_path_out is supplied seg_errors csv warnings when applicable phase input csv unless csv_path_out is supplied phase_errors csv selection and phase qc exports register input_stem _registered csv register_errors csv register_qc csv and per group logs radiomics input_stem _radiomics csv radiomics_errors csv the quickstart gives every stage a distinct output csv check error reports missing image mask paths and input output row counts a command can finish successfully even when individual volumes fail phase curation writes three additional reports input_stem _curated csv contains the best eligible ct volume per phase or mr volume per sequence phase slot in each exam use this table as pipeline input when you want representative volumes qc_ input_stem _curated csv is a wide review table with selection scores and provenance it is intended for qc rather than downstream processing qc_unresolved_phase csv retains cases unresolved before fallback with their rule model evidence for inspection selection does not reduce the main cleaned or phased table use the full table to process all candidates and select a table with the image mask paths required by the next stage selection filenames derive from the input csv even when csv_path_out names another location ingest therefore produces dicom_index_curated csv quickstart step 4 produces nifti_index_segmented_curated csv only phase offers selected_csv_path to override that location subsequent exports replace selection qc files so use separate directories for reports you want to retain see the output reference for path conventions and registration provenance resuming retries and logging parse convert segment phase register and radiomics checkpoint long runs and resume compatible work by default the parse stage of ingest also uses these controls repeat the original command with the same paths and settings to resume after interruption option behavior checkpoint_every_rows n flush checkpoint files after n processed rows checkpoint_every_sec t flush checkpoint files after t seconds no_resume ignore matching checkpoint state and start a fresh run strict_resume use content hashing for stronger input change detection retry_failed retry recorded failures while preserving successful work available for convert segment phase register and radiomics normal resume retains recorded failures after fixing their cause rerun the original command with retry_failed for example imperandi convert csv_path project tables dicom_index_clean csv output_dir project nifti csv_path_out project tables nifti_index csv retry_failed relevant input or configuration changes can invalidate cached work default fingerprints do not detect every change to image contents at the same path use strict_resume when content verification is needed more detail is in checkpoints and resume global logging options go before the subcommand imperandi log level debug log file imperandi debug log ingest root_path project dicom output_dir project tables manifest generic use imperandi command help for the options supported by your installed version and the cli reference for command behavior quality control explore patients exams and contrast phases while reviewing image slices and segmentation overlays with adjustable display settings install imperandi viewer and open a cohort table with readable nifti_path values optional mask_ _path columns provide overlays in a jupyter notebook matplotlib widget import pandas as pd from imperandi qc viewer import scanviewer cohort pd read_csv project tables nifti_index_phased csv viewer scanviewer cohort scan_col nifti_path phase_col phase constructing scanviewer displays its controls and discovers mask path columns automatically the viewer supports both ct and mr for registered outputs load nifti_index_registered csv instead for the browser viewer run from a source checkout with the viewer extra install...
|
|
| Thumbnail images (randomly selected): * Images may be subject to copyright. | |  |
|
Verified site has: 114 subpage(s). Do you want to verify them? Verify pages:
|
The site also has references to the 4 subdomain(s)
|
The site also has 11 references to external domain(s).
|
The site also has 4 references to other resources (not html/xhtml )
|
Pages verified in the last hours (randomly selected):
|
|
Top 50 hastags from of all verified websites.
| |
|
|
|
|
|
|
Load Info| page size | 68801 | | load time (s) | 1.263688 | | redirect count | 1 | | speed download | 54474 | | server IP | 140.82.121.3 |
|
|
|
|
|
|
|
|
* Image may be subject to copyright.
|
|