Echogenicity, Entropy, and Skin Temperature in Breast Cancer-Related Lymphoedema: A Cross-Sectional Study

Abstract

Background: Breast cancer–related lymphoedema (BCRL) is a frequent chronic complication associated with structural and functional changes in subcutaneous tissue. This study evaluated subcutaneous echogenicity, entropy, and thickness using ultrasound and examined their relationship with skin temperature measured by infrared thermography in breast cancer survivors. Results: Forty-five women were included (mean age 54.0 ± 7.6 years), of whom 44.4% had BCRL. After statistical adjustment, greater STT in the TB region of the homolateral limb compared with the contralateral limb remained supported, whereas other ultrasound differences were region-specific and exploratory. Participants with BCRL consistently exhibited lower Tmax across all ROIs, representing the most robust finding. Associations between ultrasound parameters and Tmax were limited and region-specific, with a statistically supported association observed only between ENT and Tmax in one forearm region. Methods: This cross-sectional exploratory study included women aged 40–70 years with a history of mastectomy. BCRL was diagnosed by indirect volumetry. Six regions of interest (ROIs) were assessed per limb (C1, C2, C3, TA in the forearm; C4 and TB in the upper arm). Ultrasound images were analysed using ImageJ version 1.54I to quantify subcutaneous tissue thickness (STT), echogenicity (ECHO), and entropy (ENT), while maximum skin temperature (Tmax) was extracted from thermographic images. Analyses accounted for repeated measurements within participants using generalized estimating equations, with adjustment for multiple comparisons. Conclusions: Ultrasonography and thermography capture partially overlapping but distinct structural and functional dimensions of tissue alteration in BCRL.

1. Introduction

Breast cancer is the most common neoplasm among women worldwide, and although its treatments are effective, they can lead to chronic complications such as secondary lymphoedema, which negatively impacts quality of life. Breast cancer-related lymphoedema (BCRL) is one of the most prevalent chronic complications following breast cancer treatment, with direct implications for skin function and subcutaneous tissue integrity [1,2]. In breast cancer, obesity, axillary lymph node dissection, radiotherapy, and chemotherapy are the main triggering factors for secondary lymphoedema [3,4].

The pathophysiology of breast cancer-related lymphoedema (BCRL) involves lymphatic stasis, vascular dysfunction, chronic inflammation, adipose tissue deposition, and fibrosis [5]. Inflammation is exacerbated by adiposity, and fibrosis, resulting from abnormal collagen deposition, contributes to disease progression. Additionally, post-injury imbalance in lymphangiogenesis further aggravates the inflammatory and immunosuppressive state [6,7,8].

Assessing the structural and functional changes associated with BCRL remains a clinical challenge. Traditional diagnostic approaches often rely on limb volume or circumference measurements, which do not capture the underlying microstructural or physiological alterations [9].

In this context, non-invasive imaging methods such as ultrasound have proven valuable in evaluating subcutaneous tissue thickness, echogenicity, and entropy, providing insights into entropy, defined as the degree of tissue disorganisation, and echogenicity, which is the ability to reflect ultrasound waves and is influenced by tissue density [10,11,12,13]. Infrared thermography has also emerged as a promising tool for detecting cutaneous temperature asymmetries linked to lymphatic dysfunction and inflammation [14].

Although both modalities have shown individual potential in the assessment of BCRL, studies combining ultrasound-based tissue characterisation with thermographic analysis remain scarce. This limits the development of integrative, multimodal approaches for understanding the local tissue changes in lymphoedema. Therefore, this study aimed to evaluate subcutaneous tissue characteristics, specifically echogenicity, entropy, and thickness, using ultrasound, and to explore their association with skin temperature assessed by infrared thermography in breast cancer survivors with and without lymphoedema.

2. Results

A total of 45 women were included in the analysis, with a mean age of 54.0 years (±7.6). Ultrasound and thermographic assessments were performed bilaterally, generating a total of 1032 region-of-interest (ROI) observations, comprising 516 ultrasound-derived and 516 thermographic-derived measurements (Figure 1). Clinical and demographic characteristics of the participants are summarised in

Table 1. Clinical characteristics of the participants (n = 45).

Variable	with BCRL (n = 20, 44.4%)	Without BCRL (n = 25, 55.6%)	p-Value
Age (years)	51.8 ± 6.8	55.7± 7.9	0.074 ¶
Body Mass Index (kg/m2)	28.8 ± 5.3	27.8 ± 5.1	0.500 ¶
Handedness			0.815 †
Right-handed	18 (90%)	23 (92%)
Left-handed	2 (10%)	2 (8%)
Type of Mastectomy			0.923 †
Total	1(5%)	2 (8%)
Modified radical	14 (70%)	17 (68%)
Radical	5 (25%)	6 (24%)
Axillary lymph node dissection	20 (100%)	23 (92%)	0.196 †
Tmáx per ROI
C1	26.3 ± 2.2	30.7 ± 1.7	<0.001 ¶
C2	26.9 ± 2.2	30.4 ± 1.4	<0.001 ¶
C3	27.05 ± 2.3	30.5 ± 1.4	<0.001 ¶
TA	26.8 ± 1.9	30.4 ± 1.4	<0.001 ¶
C4	27.5 ± 2.1	30.6 ± 1.6	<0.001 ††
TB	27.9 ± 2.1	31.2 ±1.5	<0.001 ††

^¶ Mann–Whitney Test; ^† Chi-squared; ^†† T-tests. Values expressed as mean ± standard deviation. Tmáx—Manixal Skin temperature.

.

2.1. Maximum Skin Temperature

Across all six ROIs of the homolateral limb, maximum skin temperature (Tmax) was consistently lower in participants with breast cancer-related lymphoedema (BCRL) compared with those without BCRL (Table 1). When re-analysed using generalized estimating equation (GEE) models accounting for within-subject clustering and ROI structure, these differences remained robust for all ROIs, with mean reductions in Tmax ranging from −3.18 °C to −4.37 °C in the BCRL group. All comparisons remained statistically supported after correction for multiplicity using both Holm and false discovery rate (FDR) procedures (all adjusted p < 0.001;

Table 2. Differences in maximum skin temperature (Tmax) in the homolateral limb between participants with and without BCRL (GEE models).

ROI	Mean Difference (°C)	95% CI	p-Value	p (Holm)	p (FDR)
C1	−4.37	−5.53 to −3.21	<0.001	<0.001	<0.001
C2	−3.54	−4.64 to −2.44	<0.001	<0.001	<0.001
C3	−3.45	−4.58 to −2.32	<0.001	<0.001	<0.001
TA	−3.63	−4.63 to −2.64	<0.001	<0.001	<0.001
C4	−3.18	−4.27 to −2.10	<0.001	<0.001	<0.001
TB	−3.31	−4.37 to −2.25	<0.001	<0.001	<0.001

Values represent mean differences (BCRL − no BCRL). Negative values indicate lower temperature in participants with BCRL.

).

No statistically supported differences in Tmax were observed between the homolateral and contralateral limbs after accounting for clustering and multiple comparisons (all adjusted p > 0.05).

2.2. Comparison Between Homolateral and Contralateral Limbs

Comparisons between the homolateral (HL) and contralateral (CL) limbs for ultrasound-derived parameters were re-analysed using GEE models with participant as the clustering unit and ROI as a repeated factor (

Table 3. Comparison between the homolateral limb (HL) and contralateral limb (CL) in ultrasound parameters (GEE models).

Variable	ROI	Mean Difference HL–CL (95% CI)	p-Value	p (Holm)	p (FDR)
Subcutaneous tissue thickness (STT)	C1	−0.57 (−7.18 to 6.04)	0.865	1.000	0.865
	C2	−1.10 (−6.02 to 3.82)	0.661	1.000	0.793
	C3	−1.85 (−5.85 to 2.14)	0.363	1.000	0.544
	TA	3.44 (−1.96 to 8.84)	0.211	1.000	0.493
	C4	2.66 (−1.84 to 7.15)	0.247	1.000	0.493
	TB	5.37 (2.15 to 8.59)	0.001	0.006	0.006
Entropy (ENT)	C1	−0.06 (−0.15 to 0.04)	0.239	0.801	0.359
	C2	−0.05 (−0.14 to 0.03)	0.200	0.801	0.359
	C3	0.01 (−0.08 to 0.09)	0.901	1.000	0.901
	TA	−0.02 (−0.09 to 0.04)	0.516	1.000	0.619
	C4	0.05 (0.00 to 0.11)	0.042	0.250	0.250
	TB	−0.06 (−0.13 to 0.02)	0.133	0.665	0.359
Echogenicity (ECHO)	C1	−1.66 (−6.11 to 2.79)	0.465	1.000	0.986
	C2	−0.73 (−3.63 to 2.17)	0.623	1.000	0.986
	C3	0.33 (−2.76 to 3.42)	0.834	1.000	0.986
	TA	−0.04 (−4.75 to 4.67)	0.986	1.000	0.986
	C4	0.28 (−3.30 to 3.86)	0.879	1.000	0.986
	TB	2.78 (−0.27 to 5.83)	0.074	0.443	0.443

Estimates derived from generalized estimating equations (GEE) with subject as clustering unit and exchangeable correlation structure. Values represent model-based mean differences (HL–CL) with 95% confidence intervals. p-values adjusted for multiplicity across ROIs using Holm and Benjamini–Hochberg (FDR) methods.

).

For subcutaneous tissue thickness (STT), a higher thickness in the HL compared with the CL was observed exclusively in region TB (mean difference = 5.37 pixels; 95% CI: 2.15 to 8.59). This difference remained statistically supported after correction for multiple comparisons using both Holm and FDR adjustments. No statistically supported differences were observed for STT in the remaining ROIs after multiplicity correction.

For entropy (ENT), although a higher value in the HL compared with the CL was observed in region C4 in unadjusted analyses, this difference did not remain supported after adjustment for multiple testing. No other ENT differences were observed between limbs.

Echogenicity (ECHO) did not differ between HL and CL in any ROI after correction for clustering and multiplicity.

2.3. Comparison Between Participants with and Without BCRL (Homolateral Limb)

When analyses were restricted to the homolateral limb and participants were grouped according to the presence of BCRL, GEE models revealed no statistically supported differences in subcutaneous tissue thickness or entropy across ROIs after adjustment for multiple comparisons (

Table 4. Comparison of ultrasound parameters in the homolateral limb between participants with and without BCRL (GEE models).

Variable	ROI	Mean Difference (95% CI)	p-Value	p (Holm)	p (FDR)
Subcutaneous tissue thickness (STT)	C1	6.89 (−3.39 to 17.18)	0.189	1.000	0.584
	C2	−4.53 (−13.25 to 4.20)	0.309	1.000	0.584
	C3	3.35 (−6.10 to 12.80)	0.487	1.000	0.584
	TA	3.49 (−6.05 to 13.04)	0.473	1.000	0.584
	C4	4.16 (−5.56 to 13.89)	0.401	1.000	0.584
	TB	0.45 (−8.92 to 9.82)	0.925	1.000	0.925
Entropy (ENT)	All ROIs	No differences observed after adjustment	—	—	—
Echogenicity (ECHO)	C1	6.75 (−1.71 to 15.22)	0.118	0.590	0.354
	C2	−4.33 (−13.56 to 4.90)	0.358	1.000	0.715
	C3	−0.68 (−10.86 to 9.49)	0.896	1.000	0.896
	TA	2.94 (−5.78 to 11.66)	0.509	1.000	0.763
	C4	−2.10 (−11.08 to 6.87)	0.646	1.000	0.775
	TB	9.77 (2.29 to 17.24)	0.010	0.063	0.063

GEE models with subject as clustering unit. Values represent mean differences (BCRL − no BCRL) with 95% confidence intervals. p-values adjusted for multiplicity across ROIs.

).

For echogenicity, a higher value in region TB was observed in participants with BCRL compared with those without BCRL in unadjusted analyses; however, this difference did not remain statistically supported after Holm or FDR correction. No other ROI-specific differences in echogenicity were observed.

2.4. Association Between Ultrasound Parameters and Skin Temperature

Associations between ultrasound-derived parameters and Tmax in the homolateral limb were examined using GEE models, replacing simple ROI-level correlation analyses to account for repeated measurements within subjects (

Table 5. Association between ultrasound parameters and maximum skin temperature (Tmax) in the homolateral limb (GEE models).

Predictor	ROI	β (95% CI)	p-Value	p (Holm)	p (FDR)
STT	TB	0.022 (0.004 to 0.040)	0.015	0.088	0.088
ENT	C2	0.982 (0.426 to 1.538)	<0.001	0.003	0.003
ECHO	All ROIs	No associations after adjustment	—	—	—

β coefficients represent the change in Tmax (°C) per unit increase in the ultrasound parameter. Models adjusted for ROI and clustering by subject.

).

After correction for multiplicity, entropy in region C2 demonstrated a positive association with Tmax (β = 0.98 °C per entropy unit; 95% CI: 0.43 to 1.54), which remained statistically supported following both Holm and FDR adjustments. No other associations between Tmax and subcutaneous tissue thickness or echogenicity remained statistically supported after correction for multiple testing.

3. Discussion

This study examined ultrasonographic parameters, including subcutaneous tissue thickness, echogenicity and entropy, as well as maximum skin temperature assessed by thermography, in women with and without breast cancer-related lymphoedema (BCRL), using a segmental analysis based on anatomically standardised regions of interest. After accounting for within-subject clustering and multiplicity, the findings indicate that ultrasound measures exhibit region-specific structural patterns, with differences confined to selected anatomical sites, whereas thermography consistently demonstrated lower maximum skin temperature across all regions in participants with BCRL. Associations between ultrasound parameters and skin temperature were limited and spatially heterogeneous, indicating that these modalities capture partially overlapping but distinct dimensions of tissue alteration. Together, these results highlight the complementary role of ultrasonography and thermography within an exploratory and descriptive framework, supporting the integration of both approaches for characterising structural and functional changes in breast cancer-related lymphoedema.

The complementary use of ultrasonography and thermography in this study provides clinically relevant insights by capturing distinct yet interrelated dimensions of breast cancer-related lymphoedema. Ultrasonographic parameters, such as subcutaneous tissue thickness, echogenicity, and entropy, primarily reflect structural and architectural alterations associated with tissue remodelling, fibrosis, and fluid accumulation, whereas thermography reflects functional and physiological changes related to microcirculation and inflammatory activity.

From a clinical management perspective, this dissociation reinforces the need for a multimodal assessment strategy: ultrasound may be particularly useful for identifying structural tissue changes relevant to staging, monitoring tissue composition, and guiding interventions targeting fibrosis, while thermography may support the identification of physiological alterations, treatment responsiveness, or early functional changes not yet detectable structurally. The observed regional differences, especially in the forearm and arm transition areas, further highlight the potential value of combining both modalities to inform more individualised assessment and follow-up strategies in lymphoedema care, rather than relying on a single imaging marker.

To ensure accurate and reproducible analysis, this study employed anatomically standardised segmentation of the upper limb, with ultrasound and thermographic assessments performed at seven predefined points along the anterior aspect of the arm and forearm. This methodological choice, aligned with previously established protocols, enabled topographic characterisation of localised alterations by following a consistent anatomical line from the acromion to the wrist, improving the reliability of inter-regional comparisons and enhancing the sensitivity for detecting segmental changes [12].

In the present study, increased subcutaneous tissue thickness was consistently observed in the TB region, while differences in the TA region were suggested in exploratory analyses but did not remain after statistical adjustment. This finding is partially consistent with previous research that has also reported segmental thickening in limbs affected by lymphoedema. One study identified increased skin and subcutaneous tissue thickness in specific regions of the affected limb, including the anterior distal arm, anterior and posterior forearm, and dorsum of the hand [15]. Another investigation described structural heterogeneity of the subcutaneous tissue in lymphoedema, although objective measurement of this parameter was not performed [16]. The partial divergence between studies may be due to our evaluation focusing specifically on subcutaneous tissue, rather than the dermoepidermal complex, as well as the use of anatomically standardised regions of interest (ROIs). Clinically, these findings highlight the importance of segmental assessment to detect localised changes that may be underestimated by global limb evaluation methods.

A difference in echogenicity was observed in the TB region in exploratory analyses; however, this finding should be interpreted cautiously, as it did not remain statistically supported after adjustment for multiple comparisons. This result partially differs from the literature, which more frequently reports increased echogenicity in distal anterior regions, such as the forearm and wrist, possibly associated with fibrotic remodelling and reduced tissue compliance [15]. Previous studies identifying this pattern predominantly used qualitative visual assessment, enabling the observation of ecotextural changes such as blurred fascial planes, increased brightness, and disrupted echogenic lines [13,16]. In the present study, a quantitative greyscale analysis was applied, which, although objective, may be less sensitive to subtle heterogeneity in fibrotic tissues. Detecting alterations in the TB region may reflect the influence of surgical and radiotherapy procedures that affect the proximal arm, and reinforces the need to combine qualitative and quantitative methods for a more comprehensive structural characterisation of lymphoedema.

The analysis of entropy also revealed noteworthy findings. Although increased entropy values were observed in specific regions in unadjusted analyses, only a region-specific association between entropy and skin temperature (C2) remained statistically supported after adjustment, reinforcing entropy as a potentially sensitive but spatially heterogeneous parameter. These results are consistent with the proposal of entropy as a sensitive parameter for detecting changes in oedematous and fibrotic tissues [11]. Entropy reflects the local heterogeneity of grey-scale intensity in ultrasound images and is particularly useful for identifying subtle tissue disorganisation that may not be visible through conventional echogenicity or structural measurements [11]. In the present study, the inverse correlations between entropy, echogenicity, and subcutaneous thickness support this interpretation, indicating that more compact and fibrotic regions may exhibit reduced variability and, consequently, lower entropy. Region-specific patterns observed in unadjusted analyses, including proximal regions, may relate to anatomical features and the distribution pattern of oedema, which tends to concentrate in the upper limb due to gravity and impaired lymphatic drainage. From a clinical perspective, entropy may be useful for describing tissue characteristics in areas affected by lymphoedema, especially in areas where fibrosis and tissue disorganisation predominate, and may contribute to the early identification of alterations before overt structural changes become detectable. This reinforces its potential value as a complementary imaging parameter in the multiparametric assessment of breast cancer-related lymphoedema.

Conversely, no consistent associations across regions were observed between skin temperature and ultrasonographic parameters [17], a result that contrasts with studies reporting high sensitivity of thermography in detecting thermal asymmetries between limbs [18,19,20,21]. The absence of an association in the present study likely reflects the distinct nature of the information provided by each method: thermography captures dynamic physiological responses, such as superficial perfusion and inflammation, whereas ultrasonography provides static structural information on deeper tissues, including fibrosis and fluid retention. Environmental, technical, and individual physiological factors, such as ambient temperature, humidity, acclimatisation time, and positioning, may also influence thermographic measurements [18]. These considerations reinforce that ultrasonography and thermography should not be viewed as competing methods, but rather as complementary tools.

The significant reduction in maximum skin temperature observed in all ROIs among participants with lymphoedema corroborates previous findings in the literature. A pilot study reported a negative correlation (r = −0.34) between surface temperature and the extent of secondary lymphoedema, with lower skin temperatures observed in more advanced stages of the condition [19]. This trend was attributed to progressive fibrosis of the skin and subcutaneous tissues, decreased cutaneous blood flow, and impaired heat dissipation associated with the accumulation of fibrotic and adipose tissue, which act as thermal insulators. Similarly, cold areas identified by thermography overlapped anatomically with dermal backflow (DBF) regions detected by indocyanine green lymphography, reinforcing the hypothesis that localised hypothermia reflects lymphatic dysfunction and microcirculatory compromise [22].

From a pathophysiological perspective, the decrease in skin temperature may be explained by the chronic inflammatory response, tissue sclerosis, and altered vascular dynamics seen in advanced lymphoedema. Inflammatory processes contribute to fibroblast activation and extracellular matrix deposition, which not only remodel the tissue architecture but also impair local perfusion and heat transfer. Previous studies have demonstrated the utility of thermography as a complementary diagnostic tool particularly for detecting subtle, early-stage alterations or monitoring disease progression when structural changes may not yet be apparent on ultrasound imaging [19]. In light of the present findings, thermography may also be considered a complementary approach for understanding functional tissue alterations in breast cancer-related lymphoedema, particularly when interpreted alongside structural information obtained from ultrasonography.

Taken together, the results of this study broaden the understanding that both techniques assess distinct yet complementary dimensions of lymphoedema pathophysiology. Ultrasonography, by measuring thickness, echogenicity, and entropy, provides objective information on subcutaneous tissue density and organisation, enabling the detection of changes not evident in conventional volumetric assessments [11,12]. Thermography, on the other hand, is a non-invasive and rapid method; previous studies have reported good to excellent reproducibility and sensitivity of up to 85% for detecting lymphoedema [17,23], capable of identifying thermal variations related to circulatory, inflammatory, and metabolic changes. The integration of ultrasonography and thermography may support the combined assessment of structural and functional tissue characteristics, supporting a more comprehensive descriptive evaluation of breast cancer-related lymphoedema.

Study Limitations

Despite the methodological rigour adopted in standardising image acquisition and analysis, this study presents some limitations that should be acknowledged. Subcutaneous tissue thickness was measured in pixels rather than in metric units (millimetres). Due to the absence of spatial scale information in the acquired images, it was not possible to establish a reliable post hoc conversion factor from pixels to millimetres. Although pixel-based measurements were applied consistently across all participants, this limitation reduces the direct clinical applicability of the findings and restricts comparability with studies reporting tissue thickness in standard metric units. Furthermore, the assessment of echogenicity was limited to quantitative measures, without the incorporation of validated qualitative grading systems.

Another aspect to be considered is the methodological choice of conventional ultrasonography rather than high-frequency ultrasonography. Although high-frequency systems provide superior spatial resolution and greater potential for detailing fine tissue changes, conventional ultrasonography is a validated and widely accepted modality for the characterisation of lymphoedema and represents the technology most commonly available in both clinical practice and research settings. For these reasons, conventional ultrasonography was deliberately selected, as it ensures broader applicability and external validity while still allowing for an objective and reproducible assessment of subcutaneous tissue characteristics.

This study adopts a cross-sectional design, which allows the identification of associations between imaging parameters but does not permit causal inferences. The observed relationships between ultrasonographic measures and skin temperature should therefore be interpreted as descriptive and exploratory. Consequently, conclusions regarding temporal progression or cause–effect relationships cannot be drawn from the present findings.

Future research should adopt standardised ultrasound protocols, preferably incorporating qualitative echogenicity grading and, where possible, higher-resolution technologies, in order to improve diagnostic accuracy and clinical relevance in lymphoedema assessment.

4. Materials and Methods

This is a cross-sectional study with exploratory analysis of data from a main study designed to assess the diagnostic accuracy of thermography in detecting breast cancer-related lymphoedema (BCRL), which was approved by the Research Ethics Committee for Human Subjects (approval number: 6322321) [17,23,24].

Eligible participants were women aged 40–70 years, breast cancer survivors with a history of mastectomy. Exclusion criteria included bilateral breast cancer or mastectomy, primary lymphedema, edema from other causes (rheumatologic, renal, neurologic, orthopedic, or vascular diseases), dermatological conditions (erysipelas, intertrigo, ulceration), and ongoing chemotherapy or radiotherapy. Recruitment occurred via social media, clinical partnerships, and hospital referrals. All candidates underwent initial screening at the university’s physiotherapy laboratory, followed by physical examination to confirm eligibility. Of 120 women approached, 45 met the inclusion criteria and provided written informed consent following ethical guidelines.

Data collection took place at the Women’s Health Physiotherapy Laboratory (LAFISMA) of the Federal University of Pernambuco between August 2022 and September 2023. Trained assessors conducted anamnesis (variables—age, type of mastectomy, radiotherapy, and axillary lymph node dissection), physical examination to collect body mass index, thermographic examination, and ultrasound evaluation. BCRL was diagnosed using indirect volumetry, considering lymphoedema when the volume difference between the homolateral (HL) and contralateral (CL) limbs exceeded 200 mL [25,26]. Measurements were taken with participants standing, hands against a wall, shoulders abducted to 90°, using the ulnar styloid process (C1) as reference, with circumference marks every 10 cm up to 30 cm toward the axilla [26]. Points TA and TB were located 10 cm distal and proximal to the elbow joint line, respectively; TA, C1, C2, and C3 referred to the forearm, while TB and C4 referred to the arm. (Figure 2A).

Ultrasound assessments were performed with a LOGIQ V5 device (GE HealthCare, São Paulo, Brazil) and a 7.5–10 MHz linear transducer, on the anterior segments of the forearm and arm at the same ROI sites. Participants were seated, the shoulder was positioned at 30–45°, the elbow at 45°, the forearm supinated, the wrist extended 15–30°, and with 15° ulnar deviation. Images were acquired and exported as PNG format for posterior analysis in ImageJ. Subcutaneous tissue thickness (STT) was measured in pixels using the “Straight” tool. Entropy (ENT) and Echogenicity (ECHO) were computed within polygonal ROIs delimited between two hyperechoic lines marking the subcutaneous tissue boundaries (Figure 2B). For entropy analysis, the GLCM TEXTURE plugin was used, while echogenicity was obtained through the Histogram function available in ImageJ [12,27,28,29].

Skin temperature was measured by infrared thermography. Thermal images were obtained with participants standing barefoot on a rubber mat, positioned 1 m from the FLIR C5 thermal camera (FLIR Systems, Inc., Wilsonville, OR, USA), in a controlled environment (mean temperature of 23 °C, and 50% to 55% relative humidity). Before acquisition, participants remained for 15 min with their chest, abdomen, and upper limbs exposed to ensure thermal equilibrium with the environment [17]. The regions of interest were manually delineated based on previously established anatomical landmarks. Thermal images were analyzed in FLIR Tools (emissivity 0.98, Rainbow scale, 23–37.7 °C range), with manual ROI selection over ROI marks, recording maximum skin temperatures (Tmax) [18] (Figure 2C). Maximum temperature was selected as it is considered the most appropriate parameter for assessing changes in vascular flow and inflammation [30].

The acquisition of images and measurements was performed by previously trained and calibrated researchers following a pilot calibration process (data not included). Standardised positioning and acquisition procedures were applied to minimise operator-dependent variability, including probe pressure and angle. ROI placement and feature extraction were conducted manually using predefined and standardised ROI definitions by trained assessors. Ultrasonographic analyses were performed by evaluators who were blinded to the presence or absence of lymphoedema.

Statistical analyses were conducted using JASP (v. 0.18.3.0). Continuous variables were summarised as mean ± standard deviation (SD), and categorical variables as absolute and relative frequencies. Data distribution was evaluated using the Shapiro–Wilk test.

Baseline clinical and demographic characteristics were compared between participants with and without breast cancer-related lymphoedema (BCRL) using independent t tests or Mann–Whitney tests for continuous variables, according to data distribution, and chi-squared tests for categorical variables. These analyses were performed to describe group comparability at baseline and were considered unadjusted descriptive comparisons, with two-sided p values reported.

All subsequent ROI level inferential analyses were performed using generalized estimating equations (GEE) to appropriately account for within-subject non-independence arising from repeated regions of interest (ROIs) nested within limbs and participants. GEE models were specified with a Gaussian family, an exchangeable working correlation structure, and robust (sandwich) standard errors, with clustering by participant.

For homolateral (HL) versus contralateral (CL) limb comparisons, models included limb, ROI, and limb × ROI interaction, with ROI specific contrasts (HL − CL) reported as effect estimates with 95% confidence intervals (95% CI) and corresponding p values. For between-group analyses restricted to the homolateral limb, models included BCRL group (yes or no), ROI, and group × ROI interaction, with ROI specific contrasts (BCRL − no BCRL) reported with 95% CI and p values.

Associations between ultrasound-derived parameters (STT, ECHO, ENT) and maximum skin temperature (Tmax) were examined within the same GEE framework, modelling Tmax as the dependent variable and ultrasound parameters as predictors with ROI specific effects, reporting β coefficients, 95% CI, and p values.

To control for multiple testing across the six ROIs within each outcome family, Holm and Benjamini–Hochberg false discovery rate (FDR) adjustments were applied, with adjusted p values (pHolm, pFDR) reported. All inferential conclusions presented in the Results and Discussion are based exclusively on the GEE analyses, adopting a two-sided significance level of α = 0.05.

For transparency, conventional unadjusted ROI level comparisons using paired and independent tests are provided as Supplementary Tables S1 and S2 and should be interpreted as descriptive only, as all inferential analyses are based on GEE models accounting for within-subject clustering and multiplicity.

5. Conclusions

This study demonstrates that a standardised segmental assessment of the upper limb using ultrasonography and thermography provides complementary descriptive information in the evaluation of breast cancer-related lymphoedema. After accounting for within-subject clustering and multiplicity, ultrasonography revealed region-specific structural alterations, with consistent differences in subcutaneous tissue thickness limited to selected anatomical sites, while other ultrasound-derived parameters showed spatially heterogeneous and exploratory patterns. In contrast, thermography consistently identified lower maximum skin temperature across all regions of interest in participants with lymphoedema, representing the most robust and spatially consistent finding of the present analysis.

Together, these findings indicate that ultrasonography and thermography capture partially overlapping but distinct dimensions of tissue alteration and should be interpreted as complementary tools within an exploratory framework. The use of anatomically standardised protocols supports a structured multimodal approach to the descriptive assessment of breast cancer-related lymphoedema, while highlighting the need for longitudinal and hypothesis-driven studies to clarify clinical applicability, prognostic value, and potential roles in patient monitoring and management.