Upload folder using huggingface_hub

README.md

@@ -0,0 +1,128 @@
+---
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+language:
+  - en
+tags:
+  - healthcare
+  - hearing-loss
+  - ear-disease
+  - otitis-media
+  - ent
+  - audiometry
+  - sub-saharan-africa
+  - lmic
+pretty_name: "Hearing Loss & Ear Disease (CSOM, Audiometry, Hearing Aids, ENT)"
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: ent_specialist_centre
+    data_files: data/ear_ent_specialist_centre.csv
+  - config_name: district_hospital
+    data_files: data/ear_district_hospital.csv
+    default: true
+  - config_name: rural_health_centre
+    data_files: data/ear_rural_health_centre.csv
+---
+
+# Hearing Loss & Ear Disease Dataset
+
+## Abstract
+
+This dataset provides **30,000 simulated ear disease and hearing loss records** (10,000 per scenario) from sub-Saharan Africa. Each record contains 45+ variables including ear condition, hearing assessment, risk factors, treatment, and outcomes. Three settings: ENT specialist centre (audiometry + hearing aids), district hospital (basic care), and rural health centre (minimal services).
+
+## 1. Introduction
+
+Up to 88% of hearing loss cases are preventable (WHO AFRO). CSOM is the most common ear pathology in SSA, with >56% of school children failing ear examinations. There are fewer than 1 ENT specialist per million population in most SSA countries. Audiometry is rarely available outside tertiary centres, and hearing aids reach <1% of those who need them. Untreated hearing loss in children leads to speech delay and poor school performance.
+
+**This dataset is entirely simulated. It must not be used for clinical decision-making.**
+
+## 2. Methodology
+
+### 2.1 Parameterization
+
+| Parameter | Value | Source |
+| --- | --- | --- |
+| Preventable hearing loss | Up to 88% | WHO AFRO |
+| CSOM prevalence | ~27% of cases | PMC 2017 |
+| School ear exam fail rate | 56.6% | ScienceDirect 2019 |
+| ENT specialists per million | <1 | WHO AFRO |
+| Hearing aid coverage | <1% | WHO |
+
+### 2.2 Scenario Design
+
+| Scenario | Audiometry | Hearing Aids | Surgery | ENT |
+| --- | --- | --- | --- | --- |
+| ENT specialist | Yes | Yes | Yes | Yes |
+| District hospital | No | No | No | No |
+| Rural HC | No | No | No | No |
+
+## 3. Schema
+
+| Column | Type | Description |
+| --- | --- | --- |
+| id | int | Unique identifier |
+| age | int | Patient age |
+| ear_condition | categorical | csom / aom / ome / wax / foreign_body / noise / presbycusis / congenital / ototoxicity |
+| hearing_loss_severity | categorical | normal / mild / moderate / severe / profound |
+| hearing_loss_type | categorical | conductive / sensorineural / mixed |
+| audiometry_done | binary | Audiometry performed |
+| hearing_aid_fitted | binary | Hearing aid fitted |
+| surgery_performed | binary | Ear surgery |
+| speech_development_affected | binary | Speech delay (children) |
+| school_performance_affected | binary | School impact (children) |
+| complication | binary | Complication occurred |
+
+## 4. Validation
+
+<p align="center">
+  <img src="validation_report.png" alt="Validation Report" width="100%">
+</p>
+
+Key validation checks:
+
+- **CSOM**: #1 ear condition ✓
+- **Audiometry**: 60% ENT centre → 0% elsewhere ✓
+- **Hearing aids**: 7% specialist → 0% elsewhere ✓
+- **Speech delay**: ~60% in children with moderate+ loss ✓
+- **ENT referral completion**: ~20% ✓
+
+## 5. Usage
+
+```python
+from datasets import load_dataset
+dataset = load_dataset("electricsheepafrica/hearing-loss-ear-disease", "district_hospital")
+df = dataset["train"].to_pandas()
+```
+
+## 6. Limitations
+
+- **Simulated**: Not from real ENT registries.
+- **No audiogram data**: No frequency-specific thresholds.
+- **No imaging**: No CT temporal bone.
+- **Simplified**: No detailed surgical outcomes.
+
+## 7. References
+
+1. PMC (2017). CSOM prevalence school children Malawi.
+2. Cambridge (2024). Global burden hearing impairment.
+3. Frontiers (2024). Global burden otitis media.
+4. WHO AFRO. Ear health factsheet.
+5. ScienceDirect (2019). School ear exam prevalence.
+
+## Citation
+
+```bibtex
+@dataset{esa_hearing_ear_2025,
+  title={Hearing Loss and Ear Disease Dataset},
+  author={Electric Sheep Africa},
+  year={2025},
+  publisher={Hugging Face},
+  url={https://huggingface.co/datasets/electricsheepafrica/hearing-loss-ear-disease}
+}
+```
+
+## License
+
+[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)

generate_dataset.py

@@ -0,0 +1,247 @@
+#!/usr/bin/env python3
+"""
+Literature-Informed Hearing Loss & Ear Disease Dataset
+=======================================================
+
+Generates realistic synthetic records of ear disease and hearing
+loss patients in sub-Saharan Africa, including otitis media,
+hearing assessment, hearing aids, and outcomes.
+
+References (web-searched):
+-----------
+[1]  PMC 2017. CSOM prevalence in school-age children
+     Malawi. Population-representative sample.
+[2]  Cambridge 2024. Global burden hearing impairment.
+     Highest incidence otitis media in LMICs.
+[3]  Frontiers 2024. Global burden otitis media 204
+     countries. SSA disproportionate burden.
+[4]  WHO AFRO. Up to 88% of cases preventable.
+     <1 ENT per million in most SSA countries.
+[5]  ScienceDirect 2019. 56.6% fail rate school ear
+     exams. OM with effusion most common.
+"""
+
+import numpy as np
+import pandas as pd
+import argparse
+import os
+
+SCENARIOS = {
+    'ent_specialist_centre': {
+        'description': 'ENT specialist centre with audiometry, '
+                       'tympanometry, hearing aids, surgery '
+                       '(e.g., Groote Schuur, LUTH)',
+        'audiometry_available': True,
+        'hearing_aids_available': True,
+        'surgery_available': True,
+        'ent_specialist': True,
+    },
+    'district_hospital': {
+        'description': 'District hospital with otoscopy, basic '
+                       'tuning fork tests, ear syringing, '
+                       'antibiotics (e.g., district hospitals)',
+        'audiometry_available': False,
+        'hearing_aids_available': False,
+        'surgery_available': False,
+        'ent_specialist': False,
+    },
+    'rural_health_centre': {
+        'description': 'Rural health centre, clinical diagnosis '
+                       'only, no otoscope often, traditional '
+                       'remedies common (e.g., rural DRC, Niger)',
+        'audiometry_available': False,
+        'hearing_aids_available': False,
+        'surgery_available': False,
+        'ent_specialist': False,
+    },
+}
+
+
+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}
+
+        # ── 1. Demographics ──
+        rec['age'] = max(0, min(90, int(rng.exponential(20) + 2)))
+        rec['sex'] = rng.choice(['M', 'F'], p=[0.52, 0.48])
+        rec['child'] = 1 if rec['age'] < 15 else 0
+        rec['urban'] = 1 if rng.random() < 0.35 else 0
+        rec['education'] = rng.choice(
+            ['none', 'primary', 'secondary', 'tertiary'],
+            p=[0.25, 0.35, 0.30, 0.10])
+
+        # ── 2. Ear condition ──
+        rec['ear_condition'] = rng.choice(
+            ['csom', 'aom', 'ome', 'wax_impaction', 'foreign_body',
+             'noise_induced', 'presbycusis', 'congenital', 'ototoxicity'],
+            p=[0.25, 0.15, 0.12, 0.10, 0.05,
+               0.08, 0.12, 0.05, 0.08])
+        if rec['child']:
+            rec['ear_condition'] = rng.choice(
+                ['csom', 'aom', 'ome', 'foreign_body', 'congenital', 'wax_impaction'],
+                p=[0.30, 0.25, 0.15, 0.10, 0.10, 0.10])
+        elif rec['age'] > 60:
+            rec['ear_condition'] = rng.choice(
+                ['presbycusis', 'csom', 'wax_impaction', 'noise_induced', 'ototoxicity'],
+                p=[0.40, 0.20, 0.15, 0.15, 0.10])
+
+        rec['laterality'] = rng.choice(['unilateral', 'bilateral'],
+            p=[0.55, 0.45])
+        rec['duration_months'] = max(0, min(240, int(rng.exponential(12))))
+
+        # ── 3. Risk factors ──
+        rec['recurrent_ari'] = 1 if rec['child'] and rng.random() < 0.40 else 0
+        rec['malnutrition'] = 1 if rec['child'] and rng.random() < 0.20 else 0
+        rec['overcrowding'] = 1 if rng.random() < 0.45 else 0
+        rec['noise_exposure'] = 0
+        if rec['age'] >= 15:
+            rec['noise_exposure'] = 1 if rng.random() < 0.15 else 0
+        rec['ototoxic_medication'] = 0
+        if rec['ear_condition'] == 'ototoxicity' or rng.random() < 0.05:
+            rec['ototoxic_medication'] = 1
+        rec['hiv_positive'] = 1 if rng.random() < 0.06 else 0
+        rec['traditional_ear_drops'] = 1 if rng.random() < 0.25 else 0
+
+        # ── 4. Symptoms ──
+        rec['ear_discharge'] = 0
+        if rec['ear_condition'] in ('csom', 'aom'):
+            rec['ear_discharge'] = 1 if rng.random() < 0.80 else 0
+        rec['ear_pain'] = 0
+        if rec['ear_condition'] in ('aom', 'foreign_body'):
+            rec['ear_pain'] = 1 if rng.random() < 0.85 else 0
+        rec['hearing_difficulty'] = 1 if rng.random() < 0.70 else 0
+        rec['tinnitus'] = 1 if rng.random() < 0.25 else 0
+        rec['vertigo'] = 1 if rng.random() < 0.10 else 0
+
+        # ── 5. Assessment ──
+        rec['otoscopy_done'] = 1 if rng.random() < (0.90 if sc['ent_specialist'] else 0.50) else 0
+        rec['tympanic_membrane_perforation'] = 0
+        if rec['ear_condition'] == 'csom':
+            rec['tympanic_membrane_perforation'] = 1 if rng.random() < 0.85 else 0
+
+        rec['audiometry_done'] = 0
+        if sc['audiometry_available']:
+            rec['audiometry_done'] = 1 if rng.random() < 0.60 else 0
+
+        rec['hearing_loss_severity'] = 'normal'
+        if rec['hearing_difficulty']:
+            rec['hearing_loss_severity'] = rng.choice(
+                ['mild', 'moderate', 'severe', 'profound'],
+                p=[0.35, 0.30, 0.20, 0.15])
+
+        rec['hearing_loss_type'] = 'none'
+        if rec['hearing_loss_severity'] != 'normal':
+            if rec['ear_condition'] in ('csom', 'aom', 'ome', 'wax_impaction'):
+                rec['hearing_loss_type'] = 'conductive'
+            elif rec['ear_condition'] in ('noise_induced', 'presbycusis', 'ototoxicity', 'congenital'):
+                rec['hearing_loss_type'] = 'sensorineural'
+            else:
+                rec['hearing_loss_type'] = rng.choice(
+                    ['conductive', 'sensorineural', 'mixed'], p=[0.40, 0.40, 0.20])
+
+        # ── 6. Treatment ──
+        rec['antibiotics_given'] = 0
+        if rec['ear_condition'] in ('csom', 'aom'):
+            rec['antibiotics_given'] = 1 if rng.random() < 0.70 else 0
+        rec['ear_drops_given'] = 0
+        if rec['ear_condition'] in ('csom', 'aom'):
+            rec['ear_drops_given'] = 1 if rng.random() < 0.60 else 0
+        rec['ear_syringing'] = 0
+        if rec['ear_condition'] == 'wax_impaction':
+            rec['ear_syringing'] = 1 if rng.random() < 0.60 else 0
+        rec['foreign_body_removal'] = 0
+        if rec['ear_condition'] == 'foreign_body':
+            rec['foreign_body_removal'] = 1 if rng.random() < 0.70 else 0
+
+        rec['surgery_performed'] = 0
+        if sc['surgery_available'] and rec['ear_condition'] == 'csom' and rec['tympanic_membrane_perforation']:
+            rec['surgery_performed'] = 1 if rng.random() < 0.20 else 0
+        rec['surgery_type'] = 'none'
+        if rec['surgery_performed']:
+            rec['surgery_type'] = rng.choice(
+                ['tympanoplasty', 'mastoidectomy', 'myringoplasty'],
+                p=[0.45, 0.30, 0.25])
+
+        rec['hearing_aid_fitted'] = 0
+        if sc['hearing_aids_available'] and rec['hearing_loss_severity'] in ('moderate', 'severe', 'profound'):
+            rec['hearing_aid_fitted'] = 1 if rng.random() < 0.15 else 0
+
+        rec['referred_ent'] = 0
+        if not sc['ent_specialist'] and rec['ear_condition'] in ('csom', 'congenital'):
+            rec['referred_ent'] = 1 if rng.random() < 0.30 else 0
+        rec['referral_completed'] = 0
+        if rec['referred_ent']:
+            rec['referral_completed'] = 1 if rng.random() < 0.20 else 0
+
+        # ── 7. Outcome ──
+        rec['hearing_improved'] = 0
+        if rec['ear_condition'] in ('wax_impaction', 'foreign_body', 'ome') and (
+                rec['ear_syringing'] or rec['foreign_body_removal']):
+            rec['hearing_improved'] = 1 if rng.random() < 0.80 else 0
+        elif rec['surgery_performed']:
+            rec['hearing_improved'] = 1 if rng.random() < 0.60 else 0
+        elif rec['hearing_aid_fitted']:
+            rec['hearing_improved'] = 1 if rng.random() < 0.70 else 0
+
+        rec['speech_development_affected'] = 0
+        if rec['child'] and rec['hearing_loss_severity'] in ('moderate', 'severe', 'profound'):
+            rec['speech_development_affected'] = 1 if rng.random() < 0.60 else 0
+
+        rec['school_performance_affected'] = 0
+        if rec['child'] and rec['hearing_difficulty']:
+            rec['school_performance_affected'] = 1 if rng.random() < 0.40 else 0
+
+        rec['complication'] = 0
+        if rec['ear_condition'] == 'csom' and not rec['antibiotics_given']:
+            rec['complication'] = 1 if rng.random() < 0.10 else 0
+        rec['complication_type'] = 'none'
+        if rec['complication']:
+            rec['complication_type'] = rng.choice(
+                ['mastoiditis', 'meningitis', 'brain_abscess', 'cholesteatoma'],
+                p=[0.40, 0.15, 0.10, 0.35])
+
+        records.append(rec)
+
+    df = pd.DataFrame(records)
+
+    print(f"\n{'='*65}")
+    print(f"Hearing/Ear — {scenario} (n={n}, seed={seed})")
+    print(f"{'='*65}")
+    print(f"\n  CSOM: {(df['ear_condition']=='csom').mean()*100:.1f}%")
+    print(f"  Hearing difficulty: {df['hearing_difficulty'].mean()*100:.1f}%")
+    print(f"  Audiometry done: {df['audiometry_done'].mean()*100:.1f}%")
+    print(f"  Hearing aid: {df['hearing_aid_fitted'].mean()*100:.1f}%")
+    print(f"  ENT referral: {df['referred_ent'].mean()*100:.1f}%")
+
+    return df
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        description='Generate hearing loss/ear disease 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'ear_{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'ear_{args.scenario}.csv')
+        df.to_csv(out, index=False)
+        print(f"  -> Saved to {out}")

requirements.txt

@@ -0,0 +1,3 @@
+numpy>=1.24
+pandas>=2.0
+matplotlib>=3.7

validate_dataset.py

@@ -0,0 +1,132 @@
+#!/usr/bin/env python3
+"""Validation & Diagnostic Visualization for Hearing Loss/Ear Disease Dataset."""
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+
+SCENARIOS = ['ent_specialist_centre', 'district_hospital', 'rural_health_centre']
+
+
+def load_scenarios(data_dir='data'):
+    dfs = {}
+    for sc in SCENARIOS:
+        path = os.path.join(data_dir, f'ear_{sc}.csv')
+        if os.path.exists(path):
+            dfs[sc] = pd.read_csv(path)
+    return dfs
+
+
+def make_report(dfs, output='validation_report.png'):
+    fig, axes = plt.subplots(4, 2, figsize=(16, 22))
+    fig.suptitle('Hearing Loss & Ear Disease — Validation Report',
+                 fontsize=16, fontweight='bold', y=0.98)
+    df = dfs.get('district_hospital', list(dfs.values())[0])
+    colors = ['#2ecc71', '#f39c12', '#e74c3c']
+
+    ax = axes[0, 0]
+    conds = df['ear_condition'].value_counts()
+    c_colors = ['#e74c3c', '#f39c12', '#3498db', '#9b59b6', '#2ecc71',
+                '#e67e22', '#1abc9c', '#95a5a6', '#c0392b']
+    ax.barh(range(len(conds)), conds.values, color=c_colors[:len(conds)], alpha=0.8)
+    ax.set_yticks(range(len(conds)))
+    ax.set_yticklabels([s.replace('_', ' ').title() for s in conds.index], fontsize=7)
+    ax.set_xlabel('Count')
+    ax.set_title('Ear Conditions (CSOM #1)')
+
+    ax = axes[0, 1]
+    x = np.arange(len(SCENARIOS))
+    audio = [dfs[sc]['audiometry_done'].mean()*100 for sc in SCENARIOS if sc in dfs]
+    ha = [dfs[sc]['hearing_aid_fitted'].mean()*100 for sc in SCENARIOS if sc in dfs]
+    w = 0.3
+    ax.bar(x - w/2, audio, w, label='Audiometry', color='#3498db', alpha=0.8)
+    ax.bar(x + w/2, ha, w, label='Hearing Aid', color='#2ecc71', alpha=0.8)
+    ax.set_xticks(x)
+    ax.set_xticklabels(['ENT Centre', 'District', 'Rural'], fontsize=9)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('Audiometry & Hearing Aid Access')
+    ax.legend(fontsize=8)
+
+    ax = axes[1, 0]
+    sev = df['hearing_loss_severity'].value_counts()
+    s_order = ['normal', 'mild', 'moderate', 'severe', 'profound']
+    vals = [sev.get(s, 0) for s in s_order]
+    ax.bar(range(5), vals, color=['#2ecc71', '#f1c40f', '#f39c12', '#e74c3c', '#8e44ad'], alpha=0.8)
+    ax.set_xticks(range(5))
+    ax.set_xticklabels(s_order, fontsize=9)
+    ax.set_ylabel('Count')
+    ax.set_title('Hearing Loss Severity')
+
+    ax = axes[1, 1]
+    hl_type = df[df['hearing_loss_type'] != 'none']['hearing_loss_type'].value_counts()
+    if len(hl_type) > 0:
+        ax.pie(hl_type.values,
+               labels=[s.title() for s in hl_type.index],
+               autopct='%1.0f%%', colors=['#3498db', '#e74c3c', '#f39c12'],
+               startangle=90, textprops={'fontsize': 10})
+    ax.set_title('Hearing Loss Type')
+
+    ax = axes[2, 0]
+    children = df[df['child'] == 1]
+    if len(children) > 0:
+        impacts = ['speech_development_affected', 'school_performance_affected']
+        i_labels = ['Speech Affected', 'School Affected']
+        vals = [children[i].mean()*100 for i in impacts]
+        ax.bar(range(2), vals, color=['#e74c3c', '#f39c12'], alpha=0.8)
+        ax.set_xticks(range(2))
+        ax.set_xticklabels(i_labels, fontsize=9)
+        for i, v in enumerate(vals):
+            ax.text(i, v + 1, f'{v:.0f}%', ha='center', fontsize=10)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('Impact on Children')
+
+    ax = axes[2, 1]
+    risks = ['recurrent_ari', 'malnutrition', 'overcrowding',
+             'noise_exposure', 'traditional_ear_drops']
+    r_labels = ['Recurrent ARI', 'Malnutrition', 'Overcrowding',
+                'Noise', 'Trad. Drops']
+    vals = [df[r].mean()*100 for r in risks]
+    ax.barh(range(5), vals, color='#3498db', alpha=0.7)
+    ax.set_yticks(range(5))
+    ax.set_yticklabels(r_labels, fontsize=9)
+    ax.set_xlabel('Prevalence (%)')
+    ax.set_title('Risk Factors')
+
+    ax = axes[3, 0]
+    tx = ['antibiotics_given', 'ear_drops_given', 'ear_syringing',
+          'surgery_performed', 'hearing_aid_fitted']
+    t_labels = ['Antibiotics', 'Ear Drops', 'Syringing', 'Surgery', 'Hearing Aid']
+    for i, sc_name in enumerate(SCENARIOS):
+        if sc_name in dfs:
+            d = dfs[sc_name]
+            vals = [d[t].mean()*100 for t in tx]
+            ax.plot(range(5), vals, 'o-', label=sc_name.replace('_', ' ').title()[:10],
+                    color=colors[i], linewidth=2, markersize=5)
+    ax.set_xticks(range(5))
+    ax.set_xticklabels(t_labels, fontsize=7, rotation=15)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('Treatment Cascade by Scenario')
+    ax.legend(fontsize=7)
+
+    ax = axes[3, 1]
+    ref = [dfs[sc]['referred_ent'].mean()*100 for sc in SCENARIOS if sc in dfs]
+    ref_c = [dfs[sc]['referral_completed'].mean()*100 for sc in SCENARIOS if sc in dfs]
+    ax.bar(x - w/2, ref, w, label='Referred ENT', color='#f39c12', alpha=0.8)
+    ax.bar(x + w/2, ref_c, w, label='Referral Done', color='#2ecc71', alpha=0.8)
+    ax.set_xticks(x)
+    ax.set_xticklabels(['ENT Centre', 'District', 'Rural'], fontsize=9)
+    ax.set_ylabel('Rate (%)')
+    ax.set_title('ENT Referral Gap (~20% completion)')
+    ax.legend(fontsize=8)
+
+    plt.tight_layout(rect=[0, 0, 1, 0.97])
+    plt.savefig(output, dpi=150, bbox_inches='tight')
+    print(f'Saved validation report to {output}')
+    plt.close()
+
+
+if __name__ == '__main__':
+    dfs = load_scenarios()
+    if dfs:
+        make_report(dfs)