Exoskeleton Study - Statistical Analysis Summary¶
Extracted from Informe.html (R Markdown report, Informe.Rmd).
1. Study Description¶
Comparative study evaluating the effect of a passive exoskeleton on manual load lifting biomechanics. Subjects performed lifting gestures with and without the exoskeleton (Con/Sin). Measurements cover:
- Kinovea: Angular kinematics via video analysis (4 angles x 3 gestures x 4 time instants)
- EPPA Perfil Izquierdo: Postural evaluation from left profile (cervical, trunk, shoulder, pelvic regions)
- EPPA Frente: Postural evaluation from frontal view (head/shoulder/pelvis inclination, Q-angles, rotations, distances)
- Stability: Balance platform metrics (DCL, EPE, MVL, MXE, RT) with/without exoskeleton
- CTSIB: Clinical Test of Sensory Interaction and Balance (eyes open/closed x trial conditions)
2. Methodology¶
| Aspect | Detail |
|---|---|
| Design | Within-subjects (paired), repeated measures |
| Conditions | Con exoesqueleto (with) vs Sin exoesqueleto (without) |
| Kinovea analysis | Watson-Williams test for circular data (angular comparisons) |
| EPPA analysis | Paired t-test (two-sided) |
| EPPA Frente | Paired t-test; Subject 6 excluded from analysis |
| Stability | Descriptive only (means by condition) |
| CTSIB | Descriptive only (means and SD by condition, eyes, trial) |
| Software | R with libraries: readxl, circular, DT, ggplot2, dplyr, tidyr |
| Significance level | Not explicitly stated; conventional alpha = 0.05 assumed |
3. Key Findings¶
3.1 Kinovea - Angular Kinematics¶
Descriptive Statistics (Means and SDs)¶
The Kinovea data covers 4 Angles x 3 Gestures x 4 Instants x 2 Conditions (Con/Sin), yielding 96 rows.
Angulo 1 (Angle 1) - Notable patterns: - Gesture 1, Instant 2: Con = 83.89 (SD 10.88) vs Sin = 79.05 (SD 11.36) - Gesture 2, Instant 2: Con = 101.08 (SD 10.49) vs Sin = 90.01 (SD 11.20) -- large difference - Gesture 3, Instant 2: Con = 63.43 (SD 41.23) vs Sin = 66.26 (SD 12.10) -- very high SD with exo
Angulo 2 (Angle 2) - Notable patterns: - Consistent pattern: Sin condition shows higher values at Instant 2 (e.g., 113.93 vs 99.94 in G1) - Large differences at Instant 3: Con positive vs Sin negative (e.g., G1: 2.97 vs -5.74)
Angulo 3 (Angle 3) - Lateral/frontal: - Generally small differences between conditions - Gesture 2 shows biggest gaps at Instant 2 (Con 21.68 vs Sin 35.38) and Instant 4 (Con 17.12 vs Sin 29.70)
Angulo 4 (Angle 4) - Near 90 degrees (possibly trunk/elbow): - Small and inconsistent differences between Con and Sin - Values cluster around 77-100 degrees
Watson-Williams Test Results (Gesture 2 only)¶
Test for circular mean equality between Con and Sin conditions:
| Instante | Angulo 1 | Angulo 2 | Angulo 3 | Angulo 4 |
|---|---|---|---|---|
| 1 | p = 0.61 | p = 0.35 | p = 0.43 | p = 0.69 |
| 2 | p = 0.02 | p = 0.00 | p = 0.03 | p = 0.39 |
| 3 | p = 0.17 | p = 0.00 | p = 0.59 | p = 0.79 |
| 4 | p = 0.70 | p = 0.05 | p = 0.06 | p = 0.21 |
Significant results (p < 0.05): - Angulo 1, Instante 2: p = 0.02 - Angulo 2, Instante 2: p = 0.00 - Angulo 2, Instante 3: p = 0.00 - Angulo 3, Instante 2: p = 0.03
Interpretation: The exoskeleton significantly affects joint angles primarily at Instant 2 (peak/mid-movement) across Angles 1-3 during Gesture 2. Angle 4 shows no significant effect at any instant. The effect is strongest for Angle 2 (significant at Instants 2 and 3).
3.2 EPPA - Left Profile (Perfil Izquierdo)¶
Paired t-test results comparing with/without exoskeleton:
| Region | Variable | p-value |
|---|---|---|
| C. Escapular - M. Superior | Alineacion sagital hombro [cm] | 0.22 |
| C. Escapular - M. Superior | Angulo Codo [deg] | 0.35 |
| C. Pelvica - M. Inferior | Alineacion sagital rodilla [cm] | 0.10 |
| C. Pelvica - M. Inferior | Angulo de rodilla [deg] | 0.58 |
| C. Pelvica - M. Inferior | Angulo tibio tarsiano [deg] | 0.13 |
| C. Pelvica - M. Inferior | Basculacion pelvica sagital [deg] | None (NA) |
| C. Pelvica - M. Inferior | Traslacion pelvica (Iliaco) [cm] | None (NA) |
| C. Pelvica - M. Inferior | Traslacion pelvica (Trocanter) [cm] | 0.51 |
| Cervico - Cefalica | Lordosis cervical [cm] | 0.31 |
| Cervico - Cefalica | Tangentes Occipital - Dorso [cm] | 0.30 |
| Cervico - Cefalica | Ubicacion sagital de cabeza [cm] | 0.50 |
| Tronco - Columna | Lordosis lumbar [cm] | None (NA) |
| Tronco - Columna | Tangentes Dorso - Sacro [cm] | 0.36 |
Key findings: - NO statistically significant differences found (all p > 0.05) - 3 variables returned NA/None p-values (basculacion pelvica, traslacion pelvica iliaco, lordosis lumbar) -- likely missing data or constant values - Closest to significance: Alineacion sagital rodilla (p = 0.10), Angulo tibio tarsiano (p = 0.13)
3.3 EPPA - Frontal View (Frente)¶
Subject 6 was excluded from this analysis. Paired t-test results:
| Variable | p-value |
|---|---|
| Angulo de inclinacion cabeza [deg] | 0.180 |
| Angulo de inclinacion hombros [deg] | 0.979 |
| Angulo de inclinacion Pelvis [deg] | 0.997 |
| Angulo de inclinacion Tronco [deg] | 0.054 |
| Angulo Q Derecho [deg] | None (NA) |
| Angulo Q Izquierdo [deg] | None (NA) |
| Distancia Triangulo de talla der | 0.003 |
| Distancia Triangulo de talla izq | 0.005 |
| Distancia Condilos [cm] | 0.211 |
| Distancia Gemelos [cm] | 0.714 |
| Distancia Maleolos [cm] | 0.558 |
| Rotacion Apendice Xifoides [cm] | 0.391 |
| Rotacion Eminencia Frontal Media [cm] | 0.998 |
| Rotacion Escotadura Esternal [cm] | 0.510 |
| Rotacion Espina Nasal [cm] | 0.828 |
| Rotacion Punto Mentoniano [cm] | 0.750 |
| Rotacion Sinfisis Pubiana [cm] | 0.618 |
Key findings: - Significant: Distancia Triangulo de talla derecho (p = 0.003) and izquierdo (p = 0.005) -- the exoskeleton significantly alters the waist triangle distance bilaterally - Near-significant: Angulo de inclinacion Tronco (p = 0.054) - 2 variables returned NA: Angulo Q Derecho and Izquierdo - The variable naming required extensive recoding (11 spelling variants corrected in the R code), indicating data entry inconsistencies in the source Excel
3.4 Stability¶
Descriptive means by variable and exoskeleton condition:
| Variable | Without Exo (No) | With Exo (Si) | Direction |
|---|---|---|---|
| DCL (%) | 82.69 | 84.96 | Higher with exo (+2.27) |
| EPE (%) | 77.61 | 80.04 | Higher with exo (+2.43) |
| MVL (deg/s) | 3.69 | 3.55 | Lower with exo (-0.14) |
| MXE (%) | 91.89 | 88.08 | Lower with exo (-3.81) |
| RT (s) | 0.90 | 0.87 | Lower with exo (-0.03) |
Key findings: - DCL (Directional Control Limit) and EPE (End Point Excursion) slightly improve with exo - MVL (Mean Velocity) and RT (Reaction Time) slightly decrease with exo (possibly favorable) - MXE (Maximum Excursion) decreases with exo (possibly unfavorable -- reduced range) - No inferential statistics were performed on stability data -- only descriptive
Stability data includes breakdowns by Posicion (position) and Sexo (sex) in the scatter plots, but no statistical tests were applied.
3.5 CTSIB (Clinical Test of Sensory Interaction and Balance)¶
- Bar plots and error bar plots show mean (+/- SD) of CTSIB values by:
- Exo condition (0 = without, 1 = with)
- Ojos_abiertos (eyes open/closed)
- Trial
- No inferential statistics were performed -- purely descriptive visualization
- No numerical output table was produced for CTSIB (only plots)
4. Conclusions¶
-
Kinovea/Angular: The exoskeleton significantly alters angular kinematics during Gesture 2 at the peak movement instant (Instant 2), particularly for Angles 1-3. This suggests the device modifies the lifting trajectory at mid-range.
-
EPPA Profile: No significant postural changes detected from left profile view. The exoskeleton does not appear to alter sagittal-plane postural alignment.
-
EPPA Frontal: The only significant finding is bilateral change in waist triangle distance (Distancia Triangulo de talla), suggesting the exoskeleton affects trunk geometry in the frontal plane. Trunk inclination angle approaches significance (p = 0.054).
-
Stability: Small descriptive differences suggest marginally better directional control and endpoint excursion with exo, but reduced maximum excursion. No statistical tests were applied.
-
CTSIB: Descriptive data only; no conclusions can be drawn about sensory interaction balance effects.
5. Limitations and Missing Data¶
Missing source Excel files (referenced but not present in repository)¶
Kinovea.xlsx-- Angular kinematics raw dataEPPA Cuadro comparativo perfil izquierda.xlsx-- Left profile EPPA dataEPPA Cuadro comparativo frente.xlsx-- Frontal EPPA dataEstabilidad.xlsx-- Stability platform dataCTSIB.xlsx-- CTSIB raw data
Data quality issues¶
- 3 NA p-values in EPPA Profile: Basculacion pelvica sagital, Traslacion pelvica (Iliaco), and Lordosis lumbar returned no p-value -- likely due to missing paired data or zero variance
- 2 NA p-values in EPPA Frontal: Angulo Q Derecho and Izquierdo returned no p-value
- Subject 6 excluded from EPPA Frontal analysis (no documented reason)
- 11 spelling variants of "Distancia Triangulo de talla" in the frontal EPPA Excel suggest poor data entry quality control
- Watson-Williams test applied only to Gesture 2 -- no statistical tests for Gestures 1 and 3
- Very high SD for some conditions (e.g., Angle 1, Gesture 3, Instant 2 Con: SD = 41.23 vs Sin: SD = 12.10) suggests outliers or measurement issues
- No sample size reported in the HTML (number of subjects unknown from the report alone)
- No effect sizes reported (only p-values)
- Stability and CTSIB lack inferential statistics
Encoding issues¶
- The HTML file has character encoding problems (e.g., "CC)rvico" instead of "Cervico", "C!ngulo" instead of "Angulo") suggesting the original Rmd was authored in a Latin-1 environment but rendered with incorrect encoding
6. Questions for the Clinical Team¶
-
Sample size: How many subjects participated? The Kinovea data references a "Sujeto" factor but the aggregated tables don't show N.
-
Subject 6 exclusion: Why was Subject 6 excluded from the EPPA Frontal analysis but not from other analyses?
-
Angle definitions: What do Angles 1-4 represent in the Kinovea analysis? (e.g., trunk flexion, hip angle, knee angle, elbow angle?)
-
Gesture definitions: What are Gestures 1, 2, and 3? (e.g., symmetric lift, asymmetric lift, lowering?) Why was statistical testing only performed on Gesture 2?
-
Instant definitions: What do Instants 1-4 correspond to in the lifting cycle? (e.g., start, mid-lift, peak, end?)
-
NA p-values: Why do 5 variables return no p-value? Is this due to missing measurements for certain subjects, or data recording failures?
-
Variable naming: The 11 spelling variants of "Distancia Triangulo de talla" suggest manual data entry. Has this been reviewed for other variables?
-
Stability variables: What do DCL, EPE, MVL, MXE, and RT specifically measure? What platform/device was used?
-
Missing statistical tests: Were stability and CTSIB inferential tests intentionally omitted, or should they be added?
-
Exoskeleton model: What specific exoskeleton was used? Is there a product specification?
-
EPPA Protocol: Which EPPA protocol version was followed? Was photogrammetric or goniometric measurement used?
-
Multiple comparisons: With 16 Watson-Williams tests and 30 t-tests, was any correction for multiple comparisons considered (e.g., Bonferroni, FDR)?