# PPG Dicrotic Notch Amplitude and Intraoperative Nociceptive Stimulation

## Question

Is there research on the relationship between PPG pulse-wave dicrotic notch amplitude and intraoperative noxious/nociceptive stimulation? If yes, reproduce it.

## Literature findings

Yes, related research exists, but the exact public-data situation is limited.

### Direct modern evidence

**Photoplethysmography parameters in monitoring nociception during general anesthesia**  
BMC Anesthesiology, 2025.  
URL: https://bmcanesthesiol.biomedcentral.com/articles/10.1186/s12871-025-02932-3  
PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC11831782/

Key points from the paper:

- 20 participants under general anesthesia with propofol and remifentanil.
- Four electrical stimulation intensities: 20, 40, 60, and 80 mA.
- PPG recorded from fingertip.
- Extracted PPG morphology parameters: PPGA, PBI, AA, AC, AT, MHA, MHC, MH.
- Catacrotic phase parameters, especially **AC** and **MHC**, had strong discrimination of analgesia adequacy:
  - AC: AUC = 0.851, Pk = 0.800
  - MHC: AUC = 0.837, Pk = 0.792

This is very close to the user's hypothesis because the dicrotic notch is located on the catacrotic/descending phase of the PPG waveform. However, the paper focuses on catacrotic area/mean height rather than the dicrotic notch amplitude alone.

### Older direct notch-amplitude clue

Search results also identify earlier anesthesia/nociception work mentioning **PPG dicrotic notch amplitude** as a potential nociception indicator, including Anemon Index / analgesia-quality studies. Some snippets report that combinations including **PPG dicrotic notch amplitude** helped classify movers vs non-movers under sevoflurane anesthesia. The original raw data was not found publicly.

## Reproduction limitation

The 2025 BMC electrical-stimulation PPG waveform data is not publicly available in the article or common repositories found by search. Therefore, exact reproduction of their subject-level AC/MHC results is blocked without author data.

To still test the hypothesis on open data, this reproduction uses **VitalDB v1.0.0** intraoperative PPG waveforms and treats **operation start (`opstart`)** as a proxy noxious/surgical stimulation event.

Dataset: https://physionet.org/content/vitaldb/1.0.0/

## Open-data proxy reproduction

Script:

```bash
uv run --with vitaldb --with numpy --with scipy --with pandas --with matplotlib --with scikit-learn \
  python reproductions/ppg_dicrotic_nociception/reproduce_notch_opstart.py
```

Method:

- Signal: `SNUADC/PLETH`, loaded at 100 Hz.
- Event: `opstart` from VitalDB clinical metadata.
- Pre window: `[-90, -30]` seconds before opstart.
- Post window: `[+30, +90]` seconds after opstart.
- Pulse segmentation using systolic peak detection.
- Dicrotic notch approximation: first local minimum on the post-systolic downslope.
- Extracted features per beat:
  - `PPGA = systolic_peak - pulse_foot`
  - `notch_amp = notch - pulse_foot`
  - `notch_ratio = notch_amp / PPGA`
  - `notch_depth_ratio = (systolic_peak - notch) / PPGA`
  - `IBI`
- Per case: median feature in pre and post windows.
- Statistics: Wilcoxon signed-rank test and direction-aware AUC for pre/post discrimination.

## Results

Analyzable cases: 12.

| Feature | Pre median | Post median | Median delta | Relative delta median | Wilcoxon p | Directional AUC |
|---|---:|---:|---:|---:|---:|---:|
| PPGA | 3.0411 | 3.0555 | 0.0250 | 0.0088 | 0.0522 | 0.5764 |
| notch_amp | 0.1945 | 0.2353 | 0.0187 | 0.0885 | 0.7334 | 0.5556 |
| notch_ratio | 0.0595 | 0.0725 | 0.0034 | 0.0407 | 0.8501 | 0.5417 |
| notch_depth_ratio | 0.9405 | 0.9275 | -0.0034 | -0.0036 | 0.8501 | 0.5417 |
| IBI | 0.8800 | 0.8600 | 0.0050 | 0.0063 | 0.7178 | 0.5104 |

## Interpretation

This proxy reproduction does **not** show a statistically significant dicrotic-notch response around operation start in 12 VitalDB cases. There is a small median increase in notch amplitude and notch ratio, but the effect is weak and non-significant.

This does not refute the clinical hypothesis because:

1. `opstart` is only a coarse proxy for noxious stimulation, not a controlled electrical stimulus or skin-incision timestamp.
2. VitalDB PPG quality and surgical/anesthetic conditions vary substantially.
3. Dicrotic notch detection is sensitive to waveform quality, vasopressors, sensor site, vascular tone, and filtering.
4. The stronger published 2025 result concerns catacrotic phase area/mean height, not raw notch amplitude alone.

## Files

- `reproduce_notch_opstart.py` — open-data proxy reproduction script.
- `case_features.csv` — per-case pre/post feature summary.
- `summary.json` — aggregate statistics.
- `notch_opstart_results.png` — visualization.

## Best next step

For a faithful reproduction of the 2025 BMC paper, request raw PPG waveforms and stimulation timestamps from the authors. With those data, the current pipeline can be adapted to extract:

- PPGA/PBI,
- dicrotic notch amplitude/ratio,
- AC/MHC catacrotic parameters,
- ROC AUC for 20–40 mA vs 60–80 mA,
- prediction probability Pk across stimulus intensities.