Live Intelligence · Ethiopia & Africa

Cervical cancer takes an Ethiopian woman roughly every 105 minutes — yet it is one of the few cancers that is preventable and curable when caught early. CerviScan AI combines a live pan-African data hub with AI-assisted Pap smear screening that runs entirely in your browser.

0
New cases / year · Ethiopia
0
Deaths / year · Ethiopia
Projected time until next death
1:45:00
WATCHING
0
African deaths / year
≤4%
Ethiopian women ever screened
50 / 50 FOCUS

Half of this platform is devoted to Ethiopia's national emergency; the other half tracks the whole African continent — because the solutions travel between them.

🇪🇹 ETHIOPIA
0
age-standardised cases per 100,000 women — the 2nd most common cancer among Ethiopian women. Roughly 35.2 million women aged 15+ are at risk, over 80% of cases are found late, and fewer than 1 in 25 has ever been screened.
HPV vax since 2018 VIA screening at PHCs Screening ≤4% HPV 16/18 → ~70% of cases
Source: ICO/IARC HPV Information Centre · FMOH Ethiopia
🌍 AFRICA
0
new cases across the WHO African Region every year — ≈ 1 in 4 of global cases, but ≈ 1 in 3 of global deaths, because most women are diagnosed too late.
Source: WHO AFRO
WHO TARGET 2030
90·70·90
vaccinate · screen · treat — the global elimination strategy launched in 2020.
WHO Global Strategy →
PROOF IT WORKS
>90%
HPV vaccination coverage in Rwanda's school-based programme — elimination is achievable in Africa.
WHO AFRO →
WHAT THIS PLATFORM DOES

01 · SCREEN
An EfficientNet-B0 model classifies Pap smear cells into 5 classes with 98.4% accuracy — and it runs in your browser, so patient images never leave the device.
02 · MEASURE
A living data hub of curated epidemiology from WHO, IARC/GLOBOCAN and the ICO HPV Centre — every figure carries its source, rendered into live charts.
03 · LISTEN
Live news and peer-reviewed research pulled straight from public APIs — Google News for headlines across Africa, PubMed for African science — refreshed every 10 minutes.
LIVE DATA HUB

Curated epidemiology with sources on every tile. All figures live in js/data.js — edit once, every chart and card updates.

FOCUS 1 / 2

Ethiopia MAJOR FOCUS

⏱ THE HUMAN CLOCK
0 min
the projected interval between cervical cancer deaths in Ethiopia — 0 women lost every year to a preventable disease.
Live countdown (est.)
1:45:00
Basis: ICO/IARC 2023 estimates — 5,007 deaths ÷ 525,600 min/yr
ANNUAL BURDEN
0
new cases diagnosed each year — the 2nd most common cancer among Ethiopian women.
GLOBOCAN / IARC →
AT-RISK POPULATION
0
women aged 15 and older at risk of developing cervical cancer.
ICO HPV Centre →
THE SCREENING GAP
≤4%
of eligible Ethiopian women have ever been screened — the national rate has traditionally hovered at 4% or lower.
Meta-analyses on PubMed →
NATIONAL RESPONSE
2015 →National Cancer Control Plan

Cervical cancer named a priority; VIA "screen-and-treat" rolled out to primary hospitals.

2018National HPV vaccination launch

School- and facility-based campaign starting with 14-year-old girls.

2020Global elimination pledge

Ethiopia adopts the WHO 90·70·90 strategy targets.

2022+Single-dose HPV schedule

Aligned with WHO's 2022 single-dose recommendation — more girls covered per dose.

FMOH Ethiopia →
VIRAL DRIVERS
~70%
of cervical cancers are caused by HPV types 16 & 18 — exactly what the vaccine targets. Coverage among girls 9–14 sits at 51%.
ICO →
AI OPPORTUNITY
<0.1s
for CerviScan AI to triage one cell — where a cytotechnologist may serve millions.
FOCUS 2 / 2

Africa SECONDARY FOCUS

🌍 CONTINENTAL BURDEN
0K · 0K
new cases · deaths every year in the WHO African Region. Africa carries ≈ 1 in 4 of global cases yet ≈ 1 in 3 of global deaths — the late-diagnosis penalty.
19 of top-20 highest-mortality countries are African Leading cancer killer of women in 23 African countries HIV ↑ risk ×6
Sources: WHO AFRO · IARC · WHO
ELIMINATION TRACKER · WHO 90·70·90 BY 2030
HPV vaccination coverage→ target 90%
Women screened (2× lifetime)→ target 70%
Treated after positive screen→ target 90%
Indicative regional estimates — verify at WHO AFRO
SHARE OF GLOBAL TOTAL
≈23%
of global cervical cancer cases occur in Africa
≈35%
of global cervical cancer deaths — twice the case share
IARC GLOBOCAN →
✨ WHAT SUCCESS LOOKS LIKE
>90%
Rwanda's school-based HPV vaccination coverage — proving high-income-level results are possible in Africa.
WHO AFRO →
VISUAL ANALYTICS

Rendered live with Chart.js from js/data.js. Rates are age-standardised (ASR) GLOBOCAN/ICO approximations — eastern & southern Africa carry the world's heaviest burden, while Ethiopia's population scale makes its absolute burden enormous despite a lower rate.

Incidence leaderboard · ASR per 100,000 women

Approx. GLOBOCAN/ICO estimates — verify live on GLOBOCAN →

Africa's share of the global burden

Cases share ≈ 23% Deaths share ≈ 35% Rest of world
IARC GLOBOCAN →

Projected trajectory · WHO African Region

Illustrative business-as-usual projections — IARC → · WHO →

The survival gap · deaths per 100 new cases

Derived from ICO/IARC & GLOBOCAN estimates — the late-diagnosis penalty in numbers.
PRIMARY SOURCES

WHO — Cervical CancerFact sheet: key facts, HPV, global strategy & screening guidance. IARC — GLOBOCANCancer incidence & mortality estimates for every country. ICO/IARC HPV CentreCountry fact sheets: HPV prevalence, cases, deaths, vaccination. WHO AFRORegional cervical-cancer programme & elimination updates. FMOH EthiopiaEthiopian Ministry of Health — national plans & programmes. PubMed / NCBILive research feed on this site is fetched from PubMed E-utilities.

Methodology & honesty note: Headline tiles use the most widely cited estimates (ICO/IARC 2023 for Ethiopia, WHO AFRO & GLOBOCAN for the continent) and are labelled "≈" where values vary between sources. Projections and derived ratios are illustrative. News and research are fetched live from public APIs with cached fallbacks, refreshing every 10 minutes. No number here should be cited medically without checking the linked primary source.

LIVE NEWS WIRE

Fetched live from Google News RSS — split 50/50 between Ethiopia and the rest of Africa. Auto-syncs every 10 minutes.

Synced —
Ethiopia · MAIN
Africa · Top of the Agenda
AFRICAN RESEARCH OBSERVATORY

Queried live from NCBI PubMed E-utilities — papers with African/Ethiopian author affiliations, newest first. Auto-syncs every 10 minutes.

—
CERVISCAN AI · THE ENGINE

EfficientNet-B0 transfer-learned on SIPaKMeD — the largest public Pap smear benchmark — with Grad-CAM explanations so clinicians see what the model sees.

🖼️

Input

Pap smear cell
224×224 RGB

⚗️

Preprocess

Resize · normalise
ImageNet stats

🧠

EfficientNet-B0

Transfer learning
4.3M params

📊

Classify

5-class softmax
+ confidence

🔍

Explain

Grad-CAM heatmap
nucleus attention

🚦

Confidence gate

Below 70% →
human review

🏥

Triage

Clear / review
/ urgent referral

# Input: 224 × 224 × 3 (RGB Pap smear cell)
Input → (224, 224, 3) → timm EfficientNet-B0 backbone # ImageNet pre-trained · num_classes=0
→ 1280-D feature vector → Dropout(0.4) → Dense(256) + ReLU → Dropout(0.2)
→ Dense(5) + Softmax → [Dyskeratotic · Koilocytotic · Metaplastic · Parabasal · Superficial-Intermediate]
VALIDATION RESULTS

80/20 stratified split on SIPaKMeD — 810 validation images · seed 42.

OVERALL
98.40%
Top-1 accuracy
Macro F1-score98.40%
Normal sensitivity (5-class)99.69%
Abnormal sensitivity (5-class)97.54%
PER-CLASS F1
Dyskeratotic0.9908
Koilocytotic (HPV)0.9632
Metaplastic0.9781
Parabasal1.0000
Superficial-Intermediate0.9880
🎯 BINARY TRIAGE · THE CLINICAL VIEW
99.5%
triage accuracy (Normal vs Abnormal). Abnormal sensitivity 99.4% · normal specificity 99.7%. Of 13 total errors, 10 are harmless subtype swaps caught by the review workflow — only 3 of 487 abnormal cells (0.62%) were missed.
Derived from the 5-class confusion matrix · 810 validation images
Cell typeCategoryPrecisionRecallF1Support
DyskeratoticAbnormal0.98780.99390.9908163
KoilocytoticHPV-related0.97520.95150.9632165
MetaplasticAbnormal0.97500.98110.9781159
ParabasalNormal1.00001.00001.0000157
Superficial-IntermediateNormal0.98210.99400.9880166
✅ NORMAL CELLS
99.7% sensitivity clears healthy patients correctly — reducing unnecessary referrals and clinic congestion.
⚠️ HPV CELLS
Koilocytotic recall 95.2% — still our most conservative class. HPV-flagged cells always get mandatory human review.
🚨 ABNORMAL CELLS
99.4% sensitivity at the triage level (97.5% per-subtype) — pre-cancerous change caught early.
⚕️ DISCLAIMER
Research prototype benchmarked on public data. Not a medical device — clinical validation required before deployment.
LIVE INFERENCE · RUNS IN YOUR BROWSER

· fp16 ONNX ·  MB, loaded once. Inference happens entirely on your device — no image ever leaves your machine. Research prototype, not a medical device.

Model not loaded yet…—
🔬
Drop a cell image here click, drag & drop, or paste (Ctrl+V)
✓ Pap smear cell — stained microscopy, single cell SIPaKMeD-style cytology · auto-resized to 224×224 ✗ people · objects · documents · tissue scans · screenshots

Your image appears here — it is processed locally and never uploaded.

PREDICTION
—

Confidence below 70% routes the cell to mandatory human review — the deployment safety gate. In validation, 3 of 487 abnormal cells escaped the model; the gate policy exists to catch exactly these.