Covid Net Cxr 2 Eval

---
name: covid-net-cxr-2-eval
description: Binary classification of chest X-ray images to detect SARS-CoV-2 infection. It probes a model's ability to distinguish COVID-19 positive cases from negative cases (including no pneumonia and non-SARS-CoV-2 pneumonia) using a large, multinational dataset. Use when the user wants to benchmark on COVID-Net CXR-2 benchmark dataset, or asks about evaluating this task. Reports Sensitivity.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2105.06640
  bibtex_key: [REDACTED]
  confidence: high
---

# covid-net-cxr-2-eval

> COVID-Net CXR-2: An Enhanced Deep Convolutional Neural Network Design for Detection of COVID-19 Cases from Chest X-ray Images — Pavlova et al. (2021) (arXiv:2105.06640, 2021)

## What this evaluates

Binary classification of chest X-ray images to detect SARS-CoV-2 infection. It probes a model's ability to distinguish COVID-19 positive cases from negative cases (including no pneumonia and non-SARS-CoV-2 pneumonia) using a large, multinational dataset.

## Datasets

- **COVID-Net CXR-2 benchmark dataset** — total 19203; splits: train (18803), test (400); repo http://www.covid-net.ml

## Metrics

- `Sensitivity` **(primary)** — range: percent
  - True Positives / (True Positives + False Negatives). Expressed as a percentage.
- `Positive Predictive Value (PPV)` — range: percent
  - True Positives / (True Positives + False Positives). Also known as precision.

## Input / output format

**Input**: Single chest X-ray image (preprocessed: cropped to top 8%, resized to 480×480, normalized to [0, 1]).

**Output**: Binary classification label: SARS-CoV-2 positive or SARS-CoV-2 negative.

## Scoring recipe

```python
tp = sum(1 for p, g in zip(preds, gold) if p == 1 and g == 1)
fn = sum(1 for p, g in zip(preds, gold) if p == 0 and g == 1)
fp = sum(1 for p, g in zip(preds, gold) if p == 1 and g == 0)
sensitivity = (tp / (tp + fn)) * 100
ppv = (tp / (tp + fp)) * 100
```

## Common pitfalls

- The test set is strictly balanced (200 positive, 200 negative) to ensure equal weighting of classes, unlike the imbalanced training distribution.
- Patient overlap between training and test sets must be strictly avoided to prevent data leakage, as explicitly noted in the protocol.
- The model is validated using explainability-driven methods (GSInquire) and radiologist review, which are supplementary to the standard sensitivity/PPV metrics.

## Evidence (verbatim from paper)

> The COVID-Net CXR-2 network is evaluated on a balanced test set of 200 SARS-CoV-2 positive images from 178 patients and 200 SARS-CoV-2 negative images from 100 no pneumonia and 100 non-SARS-CoV-2 pneumonia patient cases. ... the set of constraints imposed via indicator function 1_r(·) were: (1) sensitivity ≥ 95%, and (2) positive predictive value (PPV) ≥ 95%.

## Citation

```bibtex
@misc{pavlova2021covidnetcxr2,
  title={COVID-Net CXR-2: An Enhanced Deep Convolutional Neural Network Design for Detection of COVID-19 Cases from Chest X-ray Images},
  author={Pavlova et al. (2021)},
  year={2021},
  note={arXiv:2105.06640}
}
```

- arXiv: 2105.06640