5,000 construction goals
The specifications target 2,000 cases of lung cancer, 1,000 cases of enteric, gastrogen and pancreas cancer; all cases of positive tumours, with other cancers not being used to offset them.
Building specifications centred on the path of the oncologist
Priority is given to continuous time, original evidence and cross-modular linkages, rather than simply aggregating isolated tables, images or text.
Long-time-series observation window
Priority coverage has been given to the last five years, with the minimum effective time series of not less than three years in principle and containing at least a diagnosis/baseline, treatment and follow-up to multiple nodes.
ネイティブマルチモーショナルリンク
(c) Establishing traceability relationships with anonymous patient ID, consultations, examinations, sequences, specimens and group samples, and retaining the mosaics of original structures.
Compliance and project boundary
(c) Use border design according to de-marking, source clearance, authorized materials and project data; no hospitals, patients or identifiable information are made available.
An audited multi-modifiable chain of evidence for tumors
From traceable leads prior to diagnosis to pathological diagnosis, treatment process and end-of-life visits, the object can be verified by time and by connection keys.
• Retroactivity phase
Pre-diagnosis history
Priority is given to the incorporation of pre-diagnostic historical images and related reports; where there is a “comparison to the front film”, the back-image chain is created using the status tag.
ベースラインとファッスイング
Diagnosis, pathology and TNM
Retain diagnostic basis, baseline T/N/M and Stace; distinguish between available cTNM/pTTNM, date of instalment, source and version.
Treatment process
Documents, medications, surgery and images
The sequence of events surrounding the complete medical history, medical instructions, surgery/operation, examination, CT/MRI and pathological documentation of the treatment period.
04 Follow-up and outcome
Reissued, progress and last follow-up
Recurrence, progress/transfer, death, prolonged non-recurrence and final follow-up visits require time to be supported by original records without any missing extrapolation.
コアデータドメインとデリバリーオブジェクト
Specific availability is confirmed by data source permission, patient authorization/ethical material, project dictionary and batch acceptance.
| データ域 | Build Requirements and Related Objects | Research and engineering work for the future |
|---|---|---|
| Textbooks and structured events | Outpatients, hospitalization, admission/de-patientization, first course, course of illness, consultation, surgery, diagnosis, testing, medication and medical advice; full_text as original, key records are not cut off. | (c) Re-routing of patients, structuring of clinical texts, temporal sensory signs, queue definition and clinical process analysis. |
| CT/MRI and Image Report | Complete “diagnostic observations” as seen in the “inspection”, dates of examination, date of examination, date of examination, segment, period, layer; DICOM body according priority to authorization and availability. | Image time series research, image-reporting consistency verification, queue screening and multi-modular method validation. |
| Pathology, WSI and IHC | Full pathology report, specimen ID, date of extraction, specimen type, tissue and available tumor/normal properties; H& E WSI and IHC original objects confirmed by project. | Pathology - clinical linkages, blood of sample level data, proof of image methods and preparation for multiscale studies. |
| WES/WGSおよびバルクRNA | Priority FASTQ; at least reusable BAM/CRAM, VCF or expression, QC, reference version and sample metadata, clearly distinguishing between raw data and analytical results, if not available. | (b) Corporate data governance, sample linkages, cross-modular characterization studies and methodological assessments. |
| Follow-up visits, outcome and list of documents | Date of follow-up, re-entry/progress/transfer basis, last follow-up status and file_manifest; PDF, DICOM, WSI, group documents to patients and clinical clients. | Research end-of-service construction, retroactive delivery, data blood audits and quality reviews. |
Quality is not a fraction, but a set of rules that can be reviewed
Project acceptance design is co-formed with associated integrity, clinical text, time sequence logic, video/pathological metadata and batch reports.
关联完整性
Anonymized patients are identified by cross-checking the ID, consultation, examination, sequence, specimen and group samples with primary external keys and list fields to identify isolated objects and the wrong spell.
Text and Field Integrity
Full_text integrity is the core of the acceptance and inspection; the coverage and format of the records, time fields, diagnostics, tests, drugs and follow-up fields is verified.
Time sequence and clinical logic
Checking admission, examination, pathology, surgery, medical advice and follow-up time; maintaining evidence chain for comparison, re-phase and ending records of images.
Batch quality control and sample review
The quality threshold is implemented by project programme based on recommendations for data dictionary, rules check and manual sample output coverage, missing, abnormal, downgraded or eliminated.
Applicable research and data governance
Support for data preparation, methodological development, research validation and quality assessment; clinical applications still need to be independently validated by qualified institutions and professionals.
Oncology AI Method Validation
Provide a project-level data readiness framework for time perception modelling, text-structured integration, image-pathological-group linkages and model assessment.
リアルワールドリサーチチーム
The time-based retroactive basis is maintained by target population, index events, observation windows, exposure, end-of-life and quality door-control organization research queues.
Medical data infrastructure
Support for inter-team auditable synergies based on data dictionary, field mapping, humanist, self-quality self-check and batch reports.
Reference Research and Public Data Infrastructure
The resources below are used to illustrate the principles of the design of the topic and do not represent the resources that have been accessed, owned or distributed.
- Zhuang L, et al. Advancing Precision Oncology Through Modeling of Longitudinal and Multimodal Data. IEEE Reviews in Biomedical Engineering, 2026. PMC
- National Cancer Institute. Imaging Data CommonsPublic data infrastructure for the coordination of images, labelling, measurements and accompanying clinical information with DICOM.IDC
- Sivarajkumar S, et al. A multimodal generative model for structured and unstructured electronic health records. npj Health Systems, 2026. Nature
Application for the tumour data dictionary and projectization programme
Discussion with data experts on target cancer species, observation windows, research issues, data mosaics, de-labelling requirements, quality rules and project delivery boundaries, and access to data catalogues and acceptance recommendations for adaptation missions.
プロジェクト提案を入手