Morphometric Features

MyoPath extracts 37 unique morphometric features per sample from routine H&E-stained skeletal muscle sections. These features are organized into five biological categories and distilled into seven clinically interpretable pathology indicators.

Reference

Zhong H*, Gao M*, Ma S, Zhang W, Cheng N, Jiao K, Zhu B, Song J, Yan C, Yue D, Xi J, Du Z, Zhu W, Zhao C#, Luo S#. MyoPath: A deep learning pipeline for objective morphometric assessment of skeletal muscle biopsies. Journal of Pathology Informatics. 2026;100747. doi:10.1016/j.jpi.2026.100747

Seven Pathology Indicators

These seven features correspond to five pathological axes routinely assessed during muscle biopsy evaluation:

Pathological Axis	Indicator	Clinical Significance
Nuclear positioning	NCI	Centronuclear myopathy, DM1
Fiber size dysregulation	Fiber CV	Dystrophic & neurogenic processes
Fiber morphology distortion	Shape regularity	Fiber splitting, angular atrophy
Tissue replacement	Fat infiltration, Fibrosis	Late-stage dystrophy, denervation
Cellular reaction	Nuclear/muscle ratio, Inflammatory infiltration	Regeneration, inflammation

1. Nuclear Centralization Index (NCI)

Primary biomarker. Quantifies the average radial position of nuclei within muscle fibers.

$${\rho }_k=min(\frac{d_k^{\text{boundary}}}{r_i^{\text{eq}}},1.0),\text{NCI}=\frac{1}{N_{\text{analyzed}}}\sum _{k=1}^{N_{\text{analyzed}}}{\rho }_k$$

where $d_k^{\text{boundary}}$ is the distance from the nucleus centroid to the nearest fiber boundary, and $r_i^{\text{eq}}=\sqrt{a_i/\pi }$ is the equivalent circle radius.

NCI Range	Interpretation
< 0.03	Normal (subsarcolemmal nuclei)
0.03 -- 0.10	Mild centralization
0.10 -- 0.20	Moderate centralization
> 0.20	Severe centralization

Clinical evidence:

• Discriminated myopathy from controls: $p=1.3\times {10}^{-5}$, rank-biserial $r=0.69$
• DM1 showed the highest NCI (median 0.121), consistent with centronuclear pathology
• Correlated with CTG repeat count in DM1: Spearman $\rho =0.46$, $p=0.042$
• Significant dose-response trend across the GTEx myopathy spectrum (Jonckheere-Terpstra $p<{10}^{-4}$)

Nuclei are further classified into three radial zones:

Zone	Criterion	Metric
Peripheral	${\rho }_k\le 0.3$	peripheral_ratio (normal ~ 1.0)
Intermediate	$0.3<{\rho }_k<0.7$	—
Central	${\rho }_k\ge 0.7$	central_ratio (abnormal if > 0.05)

2. Fiber Size Variability Coefficient (Fiber CV)

Primary biomarker. The coefficient of variation of myofiber cross-sectional areas.

$$\text{Fiber CV}=\frac{{\sigma }_a}{\overline{a}}$$

Fiber CV Range	Interpretation
< 0.25	Normal
0.25 -- 0.40	Mild variability
0.40 -- 0.60	Moderate variability
> 0.60	Severe variability

Clinical evidence:

• Discriminated myopathy from controls: $p=2.9\times {10}^{-4}$, $r=0.58$
• Inversely correlated with grip strength in DM1: $\rho =-0.61$, $p=0.031$
• Increased with mutation severity in LGMD: 2x Missense (0.44) → LoF + Missense (0.49) → 2x LoF (0.65)
• Dimensionless and comparable across samples with different fiber calibers

3. Fiber Shape Regularity

Mean circularity (shape factor) of fiber cross-sections.

$${\text{Shape Factor}}_i=\frac{4\pi a_i}{p_i^2},\text{Mean Shape Factor}=\frac{1}{N_f}\sum _{i=1}^{N_f}{\text{Shape Factor}}_i$$

• 1.0 = perfect circle
• < 0.6 = irregular fiber morphology (splitting, angular atrophy, chronic remodeling)
• Normal range > 0.7

4. Fat Infiltration (%)

Percentage of the ROI occupied by adipose tissue.

$$\mathtt{\text{fat\_infiltration\_pct}}=\frac{A_{\text{fat}}}{A_{\text{ROI}}}\times 100\mathrm{\%}$$

• Normal range: < 5%
• A late-stage marker of dystrophic or denervation injury
• ROI-dependent: values vary with the location selected for analysis

WARNING

Fat and fibrotic replacement are nonspecific, late-stage changes that also occur with aging, disuse, and obesity. They are inherently ROI-dependent, making them less robust than NCI and fiber CV.

5. Fibrosis (%)

Connective tissue as a percentage of ROI area, computed by Boolean subtraction.

$$A_{\text{connective}}=A_{\text{ROI}}-A_{\text{muscle}}-A_{\text{fat}}$$$$\mathtt{\text{fibrosis\_pct}}=\frac{A_{\text{connective}}}{A_{\text{ROI}}}\times 100\mathrm{\%}$$

• Normal range: < 10%
• Reflects endomysial and perimysial fibrotic proliferation
• ROI-dependent

6. Nuclear/Muscle Ratio

Ratio of nuclei located within myofibers to total fiber count.

$$\mathtt{\text{nuclear\_muscle\_ratio}}=\frac{N_{\text{nuc}}^{\text{muscle}}}{N_f}$$

• Normal range: 1 -- 3
• Elevated values reflect nuclear proliferation, regeneration, or increased satellite cell activity
• Significant in HuashanMuscle cohort ($p=3.4\times {10}^{-5}$) but not in GTEx wasting spectrum

7. Inflammatory Infiltration

Nuclear density in connective tissue regions.

$$\mathtt{\text{density}}=\frac{N_{\text{nuc}}^{\text{conn}}}{A_{\text{connective}}/{10}^6}({\text{nuclei/mm}}^2)$$

• Normal range: < 2,000 nuclei/mm²
• High values suggest inflammatory cell infiltration or active fibroblast proliferation

---

Underlying Descriptive Features

The 37 features include the 7 pathology indicators above plus 30 underlying descriptive features.

Tissue Composition (10 features)

Feature	Unit	Description
roi_area_um2	µm²	Total area of the analyzed region of interest
muscle_fibers_count	count	Number of individual myofibers detected by Cellpose-SAM
muscle_area_um2	µm²	Sum of cross-sectional areas of all detected myofibers
muscle_area_pct	%	Myofiber area as fraction of ROI. Decreases with wasting, fat replacement, or fibrosis
fat_regions_count	count	Number of discrete adipose regions identified
fat_area_um2	µm²	Total adipose tissue area, excluding overlap with muscle annotations
connective_area_um2	µm²	Connective tissue area (ROI minus muscle minus fat)
connective_area_pct	%	Connective tissue as fraction of ROI
nucleus_area_um2	µm²	Total area occupied by all detected nuclei
nucleus_area_pct	%	Nuclear area as fraction of ROI. Elevated with increased cellularity

Fiber Size (7 features)

Feature	Unit	Description
fiber_mean_area_um2	µm²	Average cross-sectional area. Decreased in atrophy, increased in hypertrophy
fiber_median_um2	µm²	Median fiber area. Less sensitive to outliers
fiber_std_um2	µm²	Standard deviation of fiber areas
fiber_min_um2	µm²	Smallest detected fiber area
fiber_max_um2	µm²	Largest detected fiber area
fiber_q1_um2	µm²	25th percentile. Sensitive to grouped atrophy
fiber_q3_um2	µm²	75th percentile

Fiber Shape (3 features)

Feature	Unit	Description
shape_factor_std	dimensionless	Standard deviation of circularity across fibers
aspect_ratio_mean	dimensionless	Mean bounding-box elongation (1.0 = circular, > 2.0 = elongated)
aspect_ratio_std	dimensionless	Standard deviation of aspect ratio

Nuclear Distribution (7 features)

Feature	Unit	Description
nuclei_total_count	count	Total nuclei detected within the ROI
nuclei_in_muscle	count	Nuclei whose centroids fall within myofiber polygons
nuclei_in_connective	count	Nuclei in connective tissue regions
nuclei_unassigned	count	Nuclei not assigned to muscle or connective tissue
nuclei_per_fiber_mean	count/fiber	Average nuclei per myofiber (healthy: 1--3)
nuclei_per_fiber_std	count/fiber	Standard deviation of nuclei per fiber
nuclei_per_fiber_max	count	Maximum nuclei in any single fiber

Nuclear Localization (3 features)

Feature	Unit	Description
peripheral_ratio	0--1	Fraction of nuclei in the outer 30% of fiber radius (healthy ~ 1.0)
central_ratio	0--1	Fraction of nuclei in the inner 30% of fiber radius (abnormal if > 0.05)
multinucleated_fiber_count	count	Fibers containing more than one nucleus in cross-section

---

MyoPath Score

The MyoPath Score is a logistic regression composite of all seven pathology indicators:

$$\text{MyoPath Score}=\frac{1}{1+e^{-z}}$$$$z=-9.17+27.11\times \text{NCI}+3.96\times \text{Fiber CV}+8.22\times \text{Shape}+0.017\times \text{Fat\%}+0.043\times \text{Fibrosis\%}+0.123\times \text{NMR}+0.000001\times \text{Inflammation}$$

Performance:

• Training (GTEx): AUC = 0.788 (LOO-CV = 0.735)
• External validation (HuashanMuscle): AUC = 0.873 (without retraining)
• NCI and Fiber CV carry the largest standardized coefficients ($\beta =0.57$ and $0.53$)

Segmentation Performance

Tissue Layer	Method	Dice	IoU
Myofiber	Cellpose-SAM	0.92 ± 0.03	0.85 ± 0.06
Fat	Pixel classifier	0.95 ± 0.02	0.91 ± 0.03
Nucleus	Watershed	0.87 ± 0.04	0.78 ± 0.06
Connective tissue	Boolean subtract	0.88 ± 0.04	0.78 ± 0.06

Intraclass correlation coefficients exceed 0.88 for all seven pathology indicators.