Technical Specifications
All submitted imaging data must comply with the specifications below. Datasets that do not meet these requirements are unsuitable for processing and may be rejected or produce unreliable results.
The required MRI sequence is a diffusion-weighted, single-shot spin echo-planar imaging (EPI) acquisition with whole-brain coverage.
MRI Acquisition Parameters
Acquire whole-brain diffusion MRI using a single-shot spin echo EPI sequence sensitized to diffusion effects with the following parameters:
| Parameter | Value |
|---|---|
| Field strength, T | 3 |
| Repetition time (TR), ms | 6000–17000 |
| Echo time (TE), ms | 55–104 |
| In-plane resolution, mm² | 1.85–2.15 |
| Slice thickness, mm | 2 |
| Slices, n | 64–86 |
| Slice acquisition | Interleaved |
| Slice gap, mm | 0 |
| Flip angle, degrees | 90 |
| b-values, s/mm² | 0, 1000 |
| b0 images, n | ≥ 2 |
| Directions, n | ≥ 30 |
These acquisition requirements apply to both DICOM and NIfTI submissions. Changing file formats or metadata does not make a nonconforming acquisition suitable for processing.
Instructions for Use (IFU) Requirements
The following acquisition labels and patient information are required:
| Parameter | Value |
|---|---|
| Study Description | MRI Brain with DTI Non-Contrast |
| Series Description | neuropacs DTI |
| Age | An integer indicating the patient's age in years at the time of MRI acquisition |
| Sex (Male or Female) | The biological sex of the patient at birth |
For JSON submissions, age may be represented as a DICOM age string such as "045Y" or a whole-year string such as "45". See Patient and Examination Metadata for accepted field names and values.
Supported DICOM Transfer Syntaxes
Only lossless imaging is supported. Images acquired or previously compressed using lossy compression may fail processing or produce unreliable results. Saving a previously lossy image in an uncompressed or lossless format does not restore the discarded information.
| UID | Name |
|---|---|
1.2.840.10008.1.2 | Implicit VR Little Endian |
1.2.840.10008.1.2.1 | Explicit VR Little Endian |
1.2.840.10008.1.2.2 | Explicit VR Big Endian (Retired) |
1.2.840.10008.1.2.1.99 | Deflated Explicit VR Little Endian |
1.2.840.10008.1.2.4.70 | JPEG Lossless, Non-Hierarchical, First-Order Prediction |
1.2.840.10008.1.2.4.80 | JPEG-LS Lossless |
1.2.840.10008.1.2.4.90 | JPEG 2000 Image Compression (Lossless Only) |
1.2.840.10008.1.2.4.91 | JPEG 2000 Image Compression — accepted only when encoded losslessly |
1.2.840.10008.1.2.5 | RLE Lossless |
UID 1.2.840.10008.1.2.4.91 permits both lossless and lossy JPEG 2000 encoding. Its UID alone does not establish that an image meets the lossless-only requirement. See the DICOM JPEG 2000 transfer syntax specification.
Diffusion Parameters in DICOM
DICOM headers must retain sufficient acquisition metadata to determine the b-values and diffusion gradient directions for the complete series. Depending on the scanner and software version, these values may appear in standard attributes, sequences, or manufacturer-specific private attributes.
The following are examples of relevant attributes. A single tag or direction-count field is not sufficient to describe the complete diffusion acquisition.
| DICOM Tag | Use | Typical source |
|---|---|---|
(0008,0070) | Scanner manufacturer | Standard DICOM |
(0018,9087) | Diffusion b-value | Standard DICOM, including enhanced MR |
(0018,9089) | Diffusion gradient orientation | Standard DICOM, including enhanced MR |
(2001,1003) | Diffusion b-value | Philips |
(2005,10B0), (2005,10B1), (2005,10B2) | Diffusion gradient components | Philips |
(2005,1599) | Number of diffusion vectors in a series | Philips |
(0019,10BB), (0019,10BC), (0019,10BD) | Diffusion gradient components | GE |
(0019,10E0) | Number of diffusion directions | GE |
(0043,1039) | Diffusion b-value information | GE |
(0019,100E) | Diffusion gradient direction | Siemens |
(0019,100C) | Diffusion b-value | Siemens |
(0019,1027) | Diffusion b-matrix | Siemens |
(0019,100D) | Diffusion directionality | Siemens |
(0018,1312) | In-plane phase encoding direction (ROW or COL) | Standard DICOM |
Private attributes depend on the manufacturer, private creator, and software version. Preserve the diffusion metadata needed for conversion when deidentifying DICOM data. Refer to the converter documentation for Philips, GE, GE direction counts, and Siemens, plus the dicm2nii private attribute dictionary.
Additional DICOM Metadata
The following attributes support patient information and examination association:
| DICOM Attribute | Value |
|---|---|
Patient Age (0010,1010) | Patient age in years, for example 060Y or 048Y |
Patient Sex (0010,0040) | Patient sex, for example F or M, according to the IFU requirement above |
Accession Number (0008,0050) | Examination accession number |
Station AE Title (0008,0055) | Application Entity title recorded in this attribute, when present |
Patient Age is (0010,1010); (0010,0010) is Patient Name. Station AE Title (0008,0055) is distinct from the sender's Calling AE Title, which is supplied during DICOM network association. See the DICOM data dictionary and Calling AE Title definition.
Alternative Formats
The platform also supports the following formats for imaging ingestion:
| Format | Details |
|---|---|
| ZIP Archive (DICOM) | A single ZIP archive containing all DICOM imaging data, with an optional JSON sidecar for supported supplemental metadata |
| ZIP Archive (NIfTI) | A single ZIP archive containing NIfTI volumes (.nii or .nii.gz) with corresponding .bvec, .bval, and .json files for each scan |
For DICOM submissions, required acquisition information remains in the DICOM headers. An optional JSON sidecar does not replace missing diffusion metadata. For NIfTI submissions, use the JSON sidecar requirements below.
JSON Metadata Sidecars
A JSON sidecar supplies metadata associated with an imaging file. Acquisition metadata uses the Brain Imaging Data Structure (BIDS) conventions. Individual uploads do not require a complete BIDS dataset directory structure.
Patient and examination fields listed below are platform extensions. Their accepted aliases do not apply to BIDS acquisition field names.
File Pairing
For each NIfTI scan, provide one matching file set using the same base filename:
scan01.nii.gz
scan01.bval
scan01.bvec
scan01.json
Use a distinct base filename for each scan. Each JSON sidecar must be self-contained. When exporting from a BIDS dataset, resolve metadata inherited from higher-level JSON files into the corresponding scan sidecar before upload. See the BIDS inheritance rules.
The .bval and .bvec files must describe the same volumes, in the same order, as the NIfTI file. The .bval file contains one b-value per volume, in s/mm². The .bvec file contains three rows, with one gradient vector per column. See the BIDS diffusion file requirements.
JSON Format Requirements
- Save the sidecar as UTF-8 JSON containing a single object.
- Use double-quoted keys and strings. Comments, duplicate keys, trailing commas,
NaN, and infinity are not permitted. - Use the exact BIDS acquisition field names, capitalization, data types, and units.
- Include acquisition metadata needed to assess the MRI requirements above and all patient information required by the IFU.
- Omit unavailable optional fields. Do not substitute guessed values, empty strings,
null, or zero for missing measurements. - Use one accepted key per patient or examination field. Resolve conflicting values between files or aliases before submission.
BIDS Acquisition Metadata
Use the following conventions for acquisition metadata. The values must describe the actual acquisition.
| JSON Key | Format and units | Value or use |
|---|---|---|
MagneticFieldStrength | Number, tesla | 3 |
RepetitionTime | Number, seconds | 6.0–17.0 |
EchoTime | Number, or an applicable BIDS array of numbers, seconds | 0.055–0.104 |
FlipAngle | Number, or an applicable BIDS array of numbers, degrees | 90 |
Manufacturer | String | Scanner manufacturer, when available |
ManufacturersModelName | String | Scanner model, when available |
PhaseEncodingDirection | String: i, j, k, i-, j-, or k- | Include when required by the selected processing workflow |
TotalReadoutTime | Number, seconds | Actual effective total readout time, when required by the selected processing workflow |
Timing values in BIDS JSON use seconds. For example, TR 8500 ms is "RepetitionTime": 8.5, and TE 85 ms is "EchoTime": 0.085. Apply the acquisition limits to every applicable value. Array-valued fields must follow the applicable BIDS definition and processing requirements. See the BIDS MRI metadata definitions.
PhaseEncodingDirection describes the axes of the submitted NIfTI image. Do not infer it from an AP/PA filename label or copy a DICOM ROW/COL value into this field.
NIfTI image geometry and the .bval/.bvec files supply information such as voxel spacing, image dimensions, b-values, and gradient directions. Additional converter fields such as SliceThickness and SpacingBetweenSlices may accompany the BIDS fields; both use millimeters. SpacingBetweenSlices measures the distance between slice centers, as defined in the DICOM pixel measures specification. These are documented dcm2niix metadata fields, and their presence does not replace the acquisition requirements.
Patient and Examination Metadata
The following platform fields may be included in the same JSON object as acquisition metadata:
| Preferred JSON Key | Accepted aliases | Value | Requirement |
|---|---|---|---|
PatientAge | patientAge, patient_age | Patient age at MRI acquisition, for example "048Y" or "48" | Required |
PatientSex | patientSex, patient_sex | "F", "M", "Female", or "Male", according to the IFU requirement above | Required |
AccessionNumber | accessionNumber, accession_number | Examination accession number as a string | Optional |
Use the preferred key names for new integrations. These fields are platform extensions to the scan sidecar; full BIDS datasets describe participant information separately, commonly in participants.tsv. See the BIDS participant metadata conventions.
Keep the required study and series descriptions consistent with the source acquisition. Converter sidecars may also contain StudyDescription and SeriesDescription.
Example NIfTI Sidecar
Example scan01.json combines BIDS acquisition fields with converter metadata and the required platform patient fields:
{
"Manufacturer": "Siemens",
"MagneticFieldStrength": 3,
"RepetitionTime": 8.5,
"EchoTime": 0.085,
"FlipAngle": 90,
"PhaseEncodingDirection": "j-",
"TotalReadoutTime": 0.064,
"SliceThickness": 2,
"SpacingBetweenSlices": 2,
"StudyDescription": "MRI Brain with DTI Non-Contrast",
"SeriesDescription": "neuropacs DTI",
"PatientAge": "045Y",
"PatientSex": "M"
}
Values are illustrative and must be replaced with the actual scan metadata. This example does not establish that the accompanying images meet every acquisition requirement. Include AccessionNumber only when needed for the configured clinical workflow.
For an optional DICOM sidecar supplying patient or examination information, the supported structure remains:
{
"PatientAge": "045Y",
"PatientSex": "M",
"AccessionNumber": "ACC-12345"
}
Generating Sidecars with dcm2niix
dcm2niix generates BIDS JSON sidecars during DICOM-to-NIfTI conversion. The older dcm2nii is a separate, obsolete converter. See the dcm2niix overview.
For example, using an existing output directory:
dcm2niix -b y -ba y -z y -f "series_%s" \
-o ./output ./dicoms
| Option | Purpose |
|---|---|
-b y | Generate JSON sidecars; -b n disables them |
-ba y | Enable the converter's sidecar anonymization option |
-z y | Create gzip-compressed NIfTI output |
-f "series_%s" | Name output files using the series number |
-o ./output | Write converted files to the specified directory |
See the dcm2niix command options.
Review generated files before submission. Converter anonymization can omit patient-related fields, including age and sex, so verify that the required platform patient information has been supplied. Available acquisition metadata depends on the source DICOM files and converter support; do not invent missing values. See the dcm2niix sidecar documentation.
Privacy and Submission Checks
- Provide the required age and sex. Include an accession number only when needed for examination association in the configured workflow.
- Remove patient names, medical record numbers, dates of birth, and unnecessary identifying free text from JSON sidecars and filenames.
- Review additional converter fields before uploading. BIDS formatting and
-ba ydo not guarantee deidentification. - Confirm that each sidecar belongs to the submitted scan and that the NIfTI, b-values, and gradient vectors remain consistent after conversion or reorientation.
- Resolve missing required metadata, invalid values, conflicting information, and file mismatches before submission.
Last updated: September 8, 2026