prostate-cancer / README.md
Kossisoroyce's picture
Upload folder using huggingface_hub
f2bf984 verified
|
Raw
History Blame Contribute Delete
4.35 kB
metadata
license: cc-by-4.0
task_categories:
  - tabular-classification
language:
  - en
tags:
  - healthcare
  - prostate-cancer
  - oncology
  - psa-screening
  - late-presentation
  - sub-saharan-africa
  - lmic
pretty_name: Prostate Cancer (PSA, Staging, Treatment Access, Survival)
size_categories:
  - 10K<n<100K
configs:
  - config_name: tertiary_oncology
    data_files: data/pca_tertiary_oncology.csv
  - config_name: district_hospital
    data_files: data/pca_district_hospital.csv
    default: true
  - config_name: rural_health_centre
    data_files: data/pca_rural_health_centre.csv

Prostate Cancer Dataset

Abstract

This dataset provides 30,000 simulated prostate cancer patient records (10,000 per scenario) from sub-Saharan Africa. Each record contains 40+ variables including presentation, PSA, biopsy, staging, treatment, and survival. Three settings: tertiary oncology (40% 2yr mortality), district hospital (67%), and rural health centre (77%).

1. Introduction

Prostate cancer is the most common male cancer in SSA with high incidence and mortality rates. Late-stage presentation is the norm — most men present with advanced disease. PSA screening is inconsistent, biopsy access limited, and radiotherapy severely restricted. ADT (often surgical castration) remains the primary treatment for many. Poor survival compared to high-income countries reflects late diagnosis and treatment access barriers.

This dataset is entirely simulated. It must not be used for clinical decision-making.

2. Methodology

2.1 Parameterization

Parameter Value Source
#1 male cancer in SSA Yes PMC 2022
Late-stage presentation 55-62% Stage IV PMC 2021
PSA screening inconsistent Yes BJUI 2024
Radiotherapy access Very limited PubMed 2024
5-year survival SSA 15-55% PMC 2021

2.2 Scenario Design

Scenario PSA Biopsy RT Surgery 2yr Mort
Tertiary oncology Yes Yes Yes Yes 40%
District hospital Yes No No No 67%
Rural HC No No No No 77%

3. Schema

Column Type Description
id int Unique identifier
age int Patient age
presenting_symptom categorical luts / bone_pain / retention / haematuria
psa_tested binary PSA tested
psa_level float PSA level ng/ml
biopsy_performed binary Biopsy done
gleason_score int Gleason score
stage categorical I / II / III / IV
metastatic binary Metastatic disease
bone_metastases binary Bone mets
treatment_received categorical prostatectomy / radiotherapy / adt / palliative / none
died_within_2_years binary 2-year mortality
castration_resistant binary CRPC

4. Validation

Validation Report

Key validation checks:

  • 2yr mortality gradient: 40% → 67% → 77% ✓
  • Stage IV: 42% → 55% → 61% ✓
  • No treatment: 12% → 29% → 57% ✓
  • PSA access: 80% → 79% → 0% ✓
  • Late presentation dominant
  • Referral gap significant ✓

5. Usage

from datasets import load_dataset
dataset = load_dataset("electricsheepafrica/prostate-cancer", "district_hospital")
df = dataset["train"].to_pandas()

6. Limitations

  • Simulated: Not from real cancer registries.
  • No genomic data: No BRCA/DNA repair markers.
  • Simplified staging: No detailed TNM.
  • No longitudinal PSA: No serial measurements.
  • 2-year outcome only: No 5/10-year survival.

7. References

  1. PMC (2022). PCa screening/diagnosis/treatment in SSA.
  2. PMC (2023). Public awareness PCa screening South Africa.
  3. BJUI (2024). Global viewpoints PCa in SSA.
  4. PMC (2021). PCa survival SSA by stage.
  5. PubMed (2023). PCa management barriers in SSA.
  6. PubMed (2018). Radiotherapy PCa Ghana.
  7. PubMed (2024). Radiotherapy access barriers SSA.

Citation

@dataset{esa_prostate_cancer_2025,
  title={Prostate Cancer Dataset},
  author={Electric Sheep Africa},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/electricsheepafrica/prostate-cancer}
}

License

CC-BY-4.0