| |
| """ |
| Literature-Informed Prostate Cancer Dataset |
| ============================================= |
| |
| Generates realistic synthetic records of prostate cancer patients |
| in sub-Saharan Africa, including presentation, diagnosis, staging, |
| treatment access, and survival outcomes. |
| |
| References (web-searched): |
| ----------- |
| [1] PMC 2022. PCa screening/diagnosis/treatment in SSA. |
| Inconsistent PSA screening. Late presentation. |
| [2] PMC 2023. Public awareness PCa screening South Africa. |
| Black men disproportionately aggressive disease. |
| [3] BJUI 2024. Global viewpoints PCa in SSA. High |
| incidence and mortality. PSA not readily available. |
| [4] PMC 2021. PCa survival SSA by stage. Poor survival, |
| high proportions late stage. |
| [5] PubMed 2023. PCa management barriers in SSA. |
| Screening, diagnosis, curative Tx not available. |
| [6] PubMed 2018. Radiotherapy PCa Ghana. Emphasis on |
| curative treatment at tertiary centres. |
| [7] PubMed 2024. Radiotherapy access barriers SSA. |
| """ |
|
|
| import numpy as np |
| import pandas as pd |
| import argparse |
| import os |
|
|
| SCENARIOS = { |
| 'tertiary_oncology': { |
| 'description': 'Tertiary oncology centre with PSA, biopsy, ' |
| 'CT/MRI, radiotherapy, surgery, ADT ' |
| '(e.g., Groote Schuur, LUTH, Kenyatta)', |
| 'psa_available': True, |
| 'biopsy_available': True, |
| 'imaging_available': True, |
| 'radiotherapy_available': True, |
| 'surgery_available': True, |
| 'early_stage_proportion': 0.35, |
| 'five_year_survival': 0.55, |
| }, |
| 'district_hospital': { |
| 'description': 'District hospital with PSA, DRE, limited ' |
| 'biopsy, no radiotherapy, ADT only ' |
| '(e.g., district hospitals Uganda, Malawi)', |
| 'psa_available': True, |
| 'biopsy_available': False, |
| 'imaging_available': False, |
| 'radiotherapy_available': False, |
| 'surgery_available': False, |
| 'early_stage_proportion': 0.15, |
| 'five_year_survival': 0.30, |
| }, |
| 'rural_health_centre': { |
| 'description': 'Rural health centre, clinical diagnosis ' |
| 'only, no PSA, referral for all treatment ' |
| '(e.g., rural DRC, Niger, CAR)', |
| 'psa_available': False, |
| 'biopsy_available': False, |
| 'imaging_available': False, |
| 'radiotherapy_available': False, |
| 'surgery_available': False, |
| 'early_stage_proportion': 0.05, |
| 'five_year_survival': 0.15, |
| }, |
| } |
|
|
|
|
| def generate_dataset(n=10000, seed=42, scenario='district_hospital'): |
| rng = np.random.default_rng(seed) |
| sc = SCENARIOS[scenario] |
|
|
| records = [] |
|
|
| for idx in range(n): |
| rec = {'id': idx + 1} |
|
|
| |
| rec['age'] = max(40, min(95, int(rng.normal(67, 9)))) |
| rec['education'] = rng.choice( |
| ['none', 'primary', 'secondary', 'tertiary'], |
| p=[0.25, 0.30, 0.30, 0.15]) |
| rec['urban'] = 1 if rng.random() < 0.40 else 0 |
| rec['family_history_pca'] = 1 if rng.random() < 0.10 else 0 |
| rec['bmi'] = round(max(15, min(40, rng.normal(24, 4))), 1) |
| rec['hiv_positive'] = 1 if rng.random() < 0.06 else 0 |
| rec['diabetes'] = 1 if rng.random() < 0.08 else 0 |
| rec['hypertension'] = 1 if rng.random() < 0.30 else 0 |
|
|
| |
| rec['presenting_symptom'] = rng.choice( |
| ['luts', 'bone_pain', 'urinary_retention', 'haematuria', |
| 'weight_loss', 'incidental', 'screening'], |
| p=[0.30, 0.20, 0.15, 0.10, 0.10, 0.05, 0.10]) |
| rec['symptom_duration_months'] = max(1, min(36, int(rng.exponential(6) + 2))) |
| rec['traditional_medicine_first'] = 1 if rng.random() < 0.25 else 0 |
|
|
| |
| rec['dre_performed'] = 1 if rng.random() < 0.70 else 0 |
| rec['dre_suspicious'] = 0 |
| if rec['dre_performed']: |
| rec['dre_suspicious'] = 1 if rng.random() < 0.65 else 0 |
|
|
| rec['psa_tested'] = 0 |
| if sc['psa_available']: |
| rec['psa_tested'] = 1 if rng.random() < 0.80 else 0 |
| rec['psa_level'] = 0.0 |
| if rec['psa_tested']: |
| rec['psa_level'] = round(max(0.5, min(5000, |
| rng.lognormal(3.5, 1.5))), 1) |
| rec['psa_elevated'] = 1 if rec['psa_level'] > 4.0 else 0 |
|
|
| rec['biopsy_performed'] = 0 |
| if sc['biopsy_available'] and (rec['psa_elevated'] or rec['dre_suspicious']): |
| rec['biopsy_performed'] = 1 if rng.random() < 0.60 else 0 |
|
|
| rec['gleason_score'] = 0 |
| if rec['biopsy_performed']: |
| rec['gleason_score'] = rng.choice( |
| [6, 7, 8, 9, 10], |
| p=[0.15, 0.30, 0.25, 0.20, 0.10]) |
| rec['gleason_group'] = 0 |
| if rec['gleason_score'] == 6: |
| rec['gleason_group'] = 1 |
| elif rec['gleason_score'] == 7: |
| rec['gleason_group'] = rng.choice([2, 3]) |
| elif rec['gleason_score'] == 8: |
| rec['gleason_group'] = 4 |
| elif rec['gleason_score'] >= 9: |
| rec['gleason_group'] = 5 |
|
|
| |
| rec['stage'] = 'unknown' |
| if rng.random() < sc['early_stage_proportion']: |
| rec['stage'] = rng.choice(['I', 'II'], p=[0.30, 0.70]) |
| else: |
| rec['stage'] = rng.choice(['III', 'IV'], p=[0.35, 0.65]) |
|
|
| rec['metastatic'] = 1 if rec['stage'] == 'IV' else 0 |
| rec['bone_metastases'] = 0 |
| if rec['metastatic']: |
| rec['bone_metastases'] = 1 if rng.random() < 0.80 else 0 |
| rec['lymph_node_metastases'] = 0 |
| if rec['metastatic']: |
| rec['lymph_node_metastases'] = 1 if rng.random() < 0.40 else 0 |
|
|
| rec['ct_scan_done'] = 0 |
| if sc['imaging_available']: |
| rec['ct_scan_done'] = 1 if rng.random() < 0.50 else 0 |
| rec['bone_scan_done'] = 0 |
| if sc['imaging_available']: |
| rec['bone_scan_done'] = 1 if rng.random() < 0.30 else 0 |
|
|
| |
| rec['treatment_received'] = 'none' |
| if sc['surgery_available'] and rec['stage'] in ('I', 'II'): |
| rec['treatment_received'] = rng.choice( |
| ['radical_prostatectomy', 'radiotherapy', 'adt', 'watchful_waiting'], |
| p=[0.25, 0.25, 0.30, 0.20]) |
| elif sc['radiotherapy_available'] and rec['stage'] == 'III': |
| rec['treatment_received'] = rng.choice( |
| ['radiotherapy_adt', 'adt_only', 'none'], |
| p=[0.40, 0.45, 0.15]) |
| elif rec['stage'] == 'IV': |
| if sc['psa_available']: |
| rec['treatment_received'] = rng.choice( |
| ['adt_only', 'adt_chemo', 'palliative', 'none'], |
| p=[0.45, 0.10, 0.25, 0.20]) |
| else: |
| rec['treatment_received'] = rng.choice( |
| ['adt_only', 'palliative', 'none'], |
| p=[0.30, 0.20, 0.50]) |
| else: |
| if sc['psa_available']: |
| rec['treatment_received'] = rng.choice( |
| ['adt_only', 'palliative', 'none'], |
| p=[0.40, 0.20, 0.40]) |
| else: |
| rec['treatment_received'] = rng.choice( |
| ['palliative', 'none'], p=[0.30, 0.70]) |
|
|
| rec['adt_type'] = 'none' |
| if 'adt' in rec['treatment_received']: |
| rec['adt_type'] = rng.choice( |
| ['surgical_castration', 'lhrh_agonist', 'antiandrogen'], |
| p=[0.40, 0.35, 0.25]) |
|
|
| rec['pain_management'] = 0 |
| if rec['bone_metastases'] or rec['treatment_received'] == 'palliative': |
| rec['pain_management'] = 1 if rng.random() < 0.50 else 0 |
|
|
| rec['treatment_delay_months'] = max(0, min(24, |
| int(rng.exponential(3)))) |
| rec['referral_needed'] = 0 |
| if not sc['surgery_available'] and rec['stage'] in ('I', 'II', 'III'): |
| rec['referral_needed'] = 1 if rng.random() < 0.70 else 0 |
| rec['referral_completed'] = 0 |
| if rec['referral_needed']: |
| rec['referral_completed'] = 1 if rng.random() < 0.30 else 0 |
|
|
| |
| base_mort = 1.0 - sc['five_year_survival'] |
| if rec['stage'] == 'IV': |
| mort = base_mort * 1.3 |
| elif rec['stage'] == 'III': |
| mort = base_mort * 0.8 |
| else: |
| mort = base_mort * 0.3 |
| if rec['treatment_received'] == 'none': |
| mort *= 1.5 |
| if rec['age'] > 75: |
| mort *= 1.2 |
|
|
| rec['died_within_2_years'] = 1 if rng.random() < min(mort, 0.80) else 0 |
|
|
| rec['quality_of_life'] = rng.choice( |
| ['good', 'moderate', 'poor'], |
| p=[0.20, 0.40, 0.40] if rec['metastatic'] else [0.40, 0.40, 0.20]) |
|
|
| rec['castration_resistant'] = 0 |
| if 'adt' in rec['treatment_received']: |
| rec['castration_resistant'] = 1 if rng.random() < 0.30 else 0 |
|
|
| records.append(rec) |
|
|
| df = pd.DataFrame(records) |
|
|
| print(f"\n{'='*65}") |
| print(f"Prostate Cancer — {scenario} (n={n}, seed={seed})") |
| print(f"{'='*65}") |
| print(f"\n Stage IV: {(df['stage']=='IV').mean()*100:.1f}%") |
| print(f" PSA tested: {df['psa_tested'].mean()*100:.1f}%") |
| print(f" No treatment: {(df['treatment_received']=='none').mean()*100:.1f}%") |
| print(f" 2-yr mortality: {df['died_within_2_years'].mean()*100:.1f}%") |
|
|
| return df |
|
|
|
|
| if __name__ == '__main__': |
| parser = argparse.ArgumentParser( |
| description='Generate prostate cancer dataset') |
| parser.add_argument('--scenario', type=str, default='district_hospital', |
| choices=list(SCENARIOS.keys())) |
| parser.add_argument('--n', type=int, default=10000) |
| parser.add_argument('--seed', type=int, default=42) |
| parser.add_argument('--output', type=str, default=None) |
| parser.add_argument('--all-scenarios', action='store_true') |
| args = parser.parse_args() |
|
|
| os.makedirs('data', exist_ok=True) |
|
|
| if args.all_scenarios: |
| for sc_name in SCENARIOS: |
| df = generate_dataset(n=args.n, seed=args.seed, scenario=sc_name) |
| out = os.path.join('data', f'pca_{sc_name}.csv') |
| df.to_csv(out, index=False) |
| print(f" -> Saved to {out}\n") |
| else: |
| df = generate_dataset(n=args.n, seed=args.seed, scenario=args.scenario) |
| out = args.output or os.path.join('data', f'pca_{args.scenario}.csv') |
| df.to_csv(out, index=False) |
| print(f" -> Saved to {out}") |
|
|
