| 1 Definition | 2 Functional Requirements | 3 Syntax | 4 Semantics | 5 Conformance Testing | 6 Performance Assessment |
1 Definition
AHD Processing Types (AIH‑HPT) specifies the type of processing applied to AIH Data, independently of:
- the health data type,
- the algorithm used,
- the AIM performing the processing..
2 Functional Requirements
AIH‑HPT applies to all AIH health data objects, including but not limited to:
- BehaviouralSignalObject
- PhysiologicalSignalObject
- NeurophysiologicalSignalObject
- MedicalImagingObject
- ClinicalRecordObject
- OmicsObject
AIH‑DPT does not describe processing results, inference outcomes, or algorithms.
3 Syntax
https://schemas.mpai.community/AIH1/V1.0/data/AIHDataProcessingTypes.json
4 Semantics
| Label | Description |
| Header | AIH Data Processing Types Header – Standard “AIH‑DPT‑Vx.y”. |
| ProcessingCategory | High‑level category describing the intent of the processing applied to the data. |
| – DataPreparation | Cleaning, filtering, normalisation, alignment, segmentation. |
| – NoiseReduction | Removal or attenuation of noise or artefacts. |
| – Normalisation | Scaling or normalisation of data values. |
| – Filtering | Frequency, temporal, or spatial filtering. |
| – Segmentation | Partitioning data into meaningful units. |
| – Alignment | Temporal or spatial alignment of data elements. |
| – Registration | Spatial correspondence between data instances. |
| – FeatureExtraction | Derivation of features or representations from data. |
| – Filtering | Frequency, temporal, or spatial filtering. |
| – Segmentation | Partitioning data into meaningful units. |
| – Alignment | Temporal or spatial alignment of data elements. |
| – Registration | Spatial correspondence between data instances. |
| – RepresentationLearning | Generation of embeddings or learned representations. |
| – EmbeddingGeneration | Generation of latent vector representations. |
| – DimensionalityReduction | Reduction of representational dimensionality. |
| – InterpretationAnalysis | Pattern detection, classification, clustering, anomaly detection. |
| – Classification | Assignment of data to predefined classes. |
| – Clustering | Grouping data based on similarity. |
| – AnomalyDetection | Identification of atypical patterns. |
| – ImageSegmentation | Partitioning of images into regions of interest. |
| – PatternDetection | Detection of predefined or learned patterns. |
| – CrossModalProcessing | Correlation, fusion, or contextual integration across modalities. |
| – MultimodalFusion | Fusion of information across modalities. |
| – CrossModalCorrelation | Identification of correlations between modalities. |
| – ContextualEnrichment | Enrichment using contextual information. |
| – EstimationInference | State, risk, prognostic, or outcome estimation. |
| – StateEstimation | Estimation of system or user state. |
| – RiskEstimation | Estimation of health or outcome risks. |
| – PrognosticEstimation | Estimation of future states or outcomes. |
| – PrivacyPreservingProcessing | De‑identification, pseudonymisation, anonymisation, aggregation. |
| – DeIdentification | Removal of direct identifiers. |
| – Pseudonymisation | Replacement of identifiers with pseudonyms. |
| – Anonymisation | Irreversible removal of re‑identification potential. |
| – DataAggregation | Aggregation to reduce identifiability. |
| – KAnonymity | Application of k‑anonymity constraints. |
| – LDiversity | Application of l‑diversity constraints. |
| – DifferentialPrivacy | Application of differential privacy mechanisms. |
| – Processing Scope
|
Scope of processing, e.g. single record, session‑based, longitudinal, population‑level |
| – SingleInstance | Processing is applied to a single data instance (e.g. one signal segment, one image, one clinical record). |
| – SessionScoped | Processing is applied across multiple data instances belonging to the same interaction or session. |
| – EpisodeScoped | Processing is applied across a bounded episode of care or observation (e.g. hospital stay, treatment episode). |
| – Longitudinal | Processing is applied across temporally distributed data instances for the same subject over time. |
| – PopulationLevel | Processing is applied across data from multiple subjects or entities. |
| – CohortScoped | Processing is applied to a defined subset of a population sharing common criteria. |
| – ContextualAggregate | Processing is applied to data aggregated by contextual factors (e.g. location, activity, environment). |
| DataXMData | Information about this |
| DescrMetadata | Descriptive metadata. |
5 Conformance Testing
A Data instance Conforms with AIH Data Processing Types (AIH-ADT) if:
- Its JSON Object validates against its JSON Schema.
- Any included JSON Object validates against its JSON Schema.
- All Data in the JSON Object:
- Have the specified Data Types.
- Conform with the Qualifiers signaled in their JSON Schemas.
