---
license: bsd-3-clause
library_name: scikit-learn
pipeline_tag: tabular-regression
tags:
  - MRI
  - medical-imaging
  - ZAP
  - random-forest
  - regression
---

# ZAP Random Forest Regression Models

Trained random forest regression models for accelerated liver Z-spectrum analysis proton (ZAP) parameter estimation.

## Model configurations

Two acquisition configurations are provided:

- **12 frequency offsets:** greater acquisition acceleration
- **20 frequency offsets:** improved quantitative agreement

## Model outputs

Each configuration estimates the following voxel-wise parameters:

$$
F_f,\qquad T_{2,f}^{\mathrm{ex}},\qquad T_{2,r}^{\mathrm{ex}}.
$$

Here:

- \\(F_f\\) is the relatively free fraction.
- \\(T_{2,f}^{\mathrm{ex}}\\) is the apparent transverse exchange time of the relatively free component.
- \\(T_{2,r}^{\mathrm{ex}}\\) is the apparent transverse exchange time of the restricted component.

The restricted fraction is obtained as

$$
F_r = 1-F_f.
$$

## Model files

The release archive, `liver_zap_rfr_models_v1.zip`, contains:

- trained scikit-learn random forest regression models
- selected frequency offsets for both acquisition configurations
- model manifest
- software-environment information
- release notes
- BSD 3-Clause license
- SHA-256 integrity checksums

## Source code

The associated source code is available at [github.com/vmalis/ZAP-RFR](https://github.com/vmalis/ZAP-RFR).

## Intended use

These models are intended for research evaluation of accelerated liver ZAP imaging. They are not medical devices and are not intended for diagnosis or treatment decisions.

Application to different scanners, acquisition protocols, tissues, or populations requires independent validation.

## Security

The models use Python `joblib` serialization and therefore have Python pickle semantics. Load model files only from trusted sources and verify their SHA-256 checksums before use.