Document analysis · handwritten text recognition · bilingual clinical data

BHCP-DOC

Bilingual Handwritten Clinical Prescription Documents

A real-world dataset of 893 handwritten clinical prescriptions collected from hospitals and clinicians. The documents contain challenging English and Urdu handwriting and support research in text-line detection, handwritten text recognition, script identification, clinical document understanding, and robust bilingual OCR.

Related publication:

Bukhari, F., Coustaty, M., Luqman, M. M., Bakkali, S., Madden, M. G., & Kajla, N. I. (2026, August). BHCP-DOC: A Real-World Bilingual Dataset for Handwritten Clinical Prescription Detection and Recognition. In International Conference on Pattern Recognition (pp. 448-463). Cham: Springer Nature Switzerland.

893Prescriptions
208Clinicians
11,008Text lines
571,452Word tokens
6,356,533Characters
Dataset overview

Real-world bilingual clinical handwritten prescriptions

BHCP-DOC captures the visual and linguistic variability found in handwritten prescriptions. English and Urdu appear within the same clinical document, often alongside abbreviations, medicine names, dosage expressions, numeric measurements, symbols, and difficult or illegible writing.

Authentic clinical documents

The dataset contains 893 real-world prescription images collected from various hospitals and clinicians rather than synthetically generated handwriting.

Bilingual complexity

English and Urdu handwritten text may occur independently or within the same document, creating realistic script-switching and mixed-content recognition challenges.

Large transcription volume

The corpus contains 571,452 handwritten word tokens and 6,356,533 characters across 11,008 annotated text lines.

Dataset statistics

Stratified training, validation, and test partitions

The line-level split preserves the dataset's bilingual and heterogeneous composition for model development and evaluation.

Language composition

English
8,468 · 76.9%
Urdu
1,329 · 12.1%
Mixed/other
1,211 · 11.0%

Mixed/other lines include numeric-only, symbol-only, or illegible handwritten prescription content.

Line-level data split

PartitionLinesShare
Training8,48077.0%
Validation1,26411.5%
Test1,26411.5%
Total11,008100%
Dataset annotation

VIA JSON annotations for handwritten clinical regions

The provided annotation sample follows the VGG Image Annotator JSON structure. Rectangular regions carry language and semantic attributes that support detection, recognition, classification, and document-understanding tasks.

Current annotation attributes

Representative region attributes in the supplied VIA project include:

LanguageDosageDiagnosticSymptomsMedicine NameTextPersonal InformationNumeric Data

Schema note: the current VIA source stores the diagnostic field as Dignostic. The website uses the normalized label “Diagnostic” while retaining the original key in the downloadable example.

{
  "shape_attributes": {
    "name": "rect",
    "x": 1328,
    "y": 1295,
    "width": 147,
    "height": 95
  },
  "region_attributes": {
    "Language": "English",
    "Dosage": "",
    "Dignostic": "",
    "Symptoms": "",
    "Medicine Name": "Centrum",
    "Text": "",
    "Numeric Data": ""
  }
}
Original de-identified BHCP-DOC prescription document
Original de-identified BHCP-DOC prescription example from the dataset.
BHCP-DOC prescription displayed in the VIA annotation interface with numbered bounding boxes and semantic fields
BHCP-DOC prescription displayed in the VIA annotation interface with numbered regions and semantic annotation fields.
Sample documents

Examples of bilingual clinical handwriting

The samples demonstrate differences in layout, pen style, line spacing, clinical vocabulary, language switching, numeric expressions, and document capture conditions.

Dataset availability

Research access and release information

Clinical prescription data can contain sensitive information. The public release should use verified de-identification, institutional ethics approval, and clearly defined research-use conditions.

Repository link required

Download information

✓Version: BHCP-DOC v1.0
✓Contents: 893 documents and 11,008 handwritten text lines
✓Splits: training, validation, and test partitions

Recommended release conditions

✓Non-commercial research and educational use, subject to the final license
✓No attempt to identify patients, clinicians, or institutions
✓No direct clinical decision-making without independent validation
✓Report language-specific and mixed-content performance
Citation

Cite BHCP-DOC and the associated publication

@dataset{bukhari_bhcp_doc_2026,
  title     = {BHCP-DOC: Bilingual Handwritten Clinical Prescription Documents},
  author    = {Bukhari, Fatima and Coustaty, Mickael and Luqman, Muhammad Muzzamil and Bakkali, Souhail and Madden, Michael G. and Kajla, Nadeem Iqbal},
  year      = {2026},
  version   = {1.0},
  doi       = {10.1007/978-3-032-39379-1_31},
  url       = {https://l3i-share.univ-lr.fr/2025BHCP-DOC/#access},
  booktitle = {International Conference on Pattern Recognition (ICPR 2026 Workshops)},
  pages     = {448--463},
  year      = {2026},
  publisher   = {Springer Nature Switzerland},
  address     = {Cham}
}
Ethics, license, and contact

Final publication details

Complete before release

Ethics approval

This dataset was approved by the Faculty of Computing, MNS University of Agriculture, Multan, Pakistan. The study used retrospective clinical prescription images. Patient consent was waived because all records were fully anonymized before analysis. Personal identifiers including patient names, addresses, phone numbers, hospital IDs, physician signatures, QR codes, barcodes, and any other identifiable information were removed or masked before inclusion in the dataset.

Complete before release

License

The dataset is provided solely for non-commercial scientific research. Users must not attempt to re-identify individuals or redistribute the dataset without permission. Commercial use requires written authorization from the dataset owners.

Complete before release

Contact

In case of any problem, please contact
"l3i-pn$@$univ$-$lr$.$fr" (without $)