# Outline v3 (FINAL structure — proposal A) — Role of AI in the Analysis of Unstructured Clinical Databases

Deck: `deck/index.html` (Reveal.js, 15 slides, fonts embedded).
Presenters: **MP** = Dr. Marco Pancotti · **SP** = Dr.ssa Sara Paratico.

| #  | Slide                                          | Master       | By | Time | Status |
|----|------------------------------------------------|--------------|----|------|--------|
| 01 | Title                                          | title        | MP | 40s  | ✅ final |
| 02 | Where we started (4 systems + 2 missing)       | interactive  | MP | 65s  | ✅ final |
| 03 | The project at a glance (flow + brain popups)  | architecture | MP | 75s  | ✅ final |
| 04 | What the text miner reads (numbers)            | stats        | SP | 60s  | ✅ real data |
| 05 | The hard problems of clinical NLP              | concept      | SP | 75s  | 🟡 draft |
| 06 | Building the clinical ontology                 | concept      | SP | 75s  | 🟡 draft |
| 07 | Trust the text: validation & human review      | concept      | SP | 75s  | 🟡 draft |
| 08 | AI as co-engineer (mapping YAMLs)              | artifact     | MP | 70s  | ✅ real |
| 09 | Genetic data — many sources, one patient       | architecture | MP | 60s  | ⚠️ wording TBD (a/b) |
| 10 | ThothII: ask the warehouse in plain English    | concept      | MP | 75s  | 🟡 draft |
| 11 | From datamarts to predictive statistics & ML   | concept      | MP | 60s  | 🟡 draft |
| 12 | Live demo cue (star schema · Superset · ThothII)| divider     | MP | ~2min live | ✅ |
| 13 | Lessons learned                                | concept      | MP | 60s  | 🟡 draft |
| 14 | (merged into 15)                               | —            | —  | —    | — |
| 15 | Thank you + Q&A (+ Substack pointer)           | closing      | MP | 25s  | 🟡 draft |

Talk ≈ 13.4 min + ~2 min demo. 15 slides = user cap reached.

## Confirmed decisions
- Project named **AritmoLab** everywhere (internal codename never appears).
- Slide 02 (Where we started): map of legacy systems — Cardioref / Genetic data /
  Omics Portal / ECG subsystems, each clickable → popup (usage + problem box);
  two dashed "MISSING" cards (Health Intelligence, ML-ready datamarts).
- Slide 03 flow: 3 sources + "Future sources" (dashed) → staging → integration 🧠
  → star schema → datamarts 🧠 → AritmoLab (DWH + Portal); brain popups =
  The mappings / Reading the clinical text / Datamarts from plain English.
- Text-analysis block (SP): slides 04–07. Scripts to be validated by SP.
- Speaker view (S): Tall layout customized — upcoming 25%, timer bottom-left,
  notes 1.45em. Default: upcoming 20%, notes 80%.
- Fonts embedded (Source Sans 3 + Source Serif 4, OFL) — no Google dependency.

## OPEN
- Slide 09 genetics wording: (a) deterministic reconciliation [recommended] vs
  (b) AI upstream (only if MP confirms it happens outside the ETL).
- Draft slides 05-07, 10-11, 13-15: review one by one.
- Dark theme of the whole deck: final pass.
- "twenty years" (slide 01 script + slide 02): confirm real data span.