rphammonds/scalesurfer-stats-v8

ScaleSurfer FreeSurfer v8 Stats Prediction Model

This repository contains a ScaleSurfer multi-head model for predicting FreeSurfer-style .stats targets from a T1w image and an aparc+aseg segmentation.

Files

• stats_model.pt: full PyTorch stats checkpoint, including frozen encoder weights, prediction heads, target normalization, and target columns.
• config.json: architecture and feature-schema metadata needed by ScaleSurferStatsPredictor.
• metadata.json: checksums and training/evaluation metadata.
• summary.csv, history.csv, target_metrics.csv: copied training diagnostics.

Test Summary

Group	Targets	Values	Normalized MAE	Median absolute percent error
aseg	116	4734	0.684679242208474	8.691529750823975
global	22	1056	0.22105189112798226	1.9054512977600098
lh_aparc	306	10404	0.4361358032305286	7.891855001449585
rh_aparc	306	10404	0.44021236654629725	7.936147093772888

Loading

from scalesurfer.stats import ScaleSurferStatsPredictor

predictor = ScaleSurferStatsPredictor.from_pretrained(8)
features = predictor.predict_subjects(subjects_dir, subjects, return_format="wide")

This model is intended for research workflows and is not a clinical diagnostic device.