Correct internal test set size to 189 patients

README.md

@@ -15,13 +15,13 @@ tags:
 
 **Sequence-agnostic whole-liver segmentation on abdominal MRI.**
 
-LiverSegMRI is a five-fold nnU-Net v2 ensemble (`3d_fullres`, `nnUNetTrainerNoMirroring`) that segments the whole 3D liver on a single abdominal MRI sequence. It is sequence agnostic and handles:
+LiverSegMRI is a five-fold nnU-Net v2 ensemble (`3d_fullres`, `nnUNetTrainerNoMirroring`) that segments the whole liver on a single abdominal MRI sequence. It needs no sequence labels and handles:
 
 - T1-weighted in- and opposed-phase
-- T1W pre-contrast and dynamic post-contrast (early and late arterial, portal venous, transitional or delayed)
-- T1W hepatobiliary phase
+- pre-contrast and dynamic post-contrast (early and late arterial, portal venous, transitional or delayed)
+- hepatobiliary phase
 - T2-weighted
-- Diffusion-weighted and ADC
+- diffusion-weighted and ADC
 
 | Resource | Link |
 |---|---|
@@ -89,7 +89,7 @@ Patient-wise mean Dice:
 | Test set | Patients | LiverSegMRI | TotalSegmentator MRI | MRAnnotator |
 |---|---:|---:|---:|---:|
 | External: Duke Liver Dataset and CirrMRI600+ | 487 | **0.944** | 0.886 | 0.889 |
-| Internal: held out | 188 | **0.985** | 0.880 | 0.894 |
+| Internal: held out | 189 | **0.985** | 0.880 | 0.894 |
 | Development: out-of-fold | 1,058 | 0.983 | — | — |
 
 External test set, LiverSegMRI vs comparators:
@@ -99,7 +99,7 @@ External test set, LiverSegMRI vs comparators:
 | HD95 (mm) | 6.6 | 15.3 | 27.0 |
 | Volume error (%) | 5.0 | 13.9 | 12.5 |
 
-- **By field strength:** performance was similar at 1.5 T and 3.0 T.
+- **By field strength:** in Mount Sinai patients with DICOM metadata, performance was similar at 1.5 T and 3.0 T.
 - **Comparisons with the other models:** all paired comparisons were significant with Holm-adjusted P < .001.
 
 ### Performance across sequence types and liver morphology
@@ -122,6 +122,8 @@ Three examinations, one per row, each shown with the manual reference (red) and
 - **E–H, T2-weighted imaging.** The liver is dark against a bright spleen; both comparators under-segment the dome and the left lobe (0.96 vs 0.74 and 0.75).
 - **I–L, diffusion-weighted imaging.** The lowest-resolution sequence shown: TotalSegmentator MRI misses the posterior right lobe and MRAnnotator fragments inside the parenchyma, while LiverSegMRI reproduces the reference almost exactly (0.98 vs 0.68 and 0.79).
 
+The pattern matches the quantitative result above: the models agree most on high-resolution contrast-enhanced T1-weighted imaging and diverge on T2-weighted and diffusion-weighted acquisitions.
+
 See the paper for sequence-wise, subgroup, failure, and label-matched analyses.
 
 ## Intended use and limitations