Comparison of a novel algorithm quantitatively estimating epifascial fibrosis in three-dimensional computed tomography images to other clinical lymphedema grading methods

Abstract

No method has yet been approved for detecting lymphedema fibrosis before its progression. This study assessed the feasibility of computed tomography-based estimation of fibrosis. This observational, cross-sectional study included patients with lymphedema affecting one limb. Three types (maximum, mean, minimum) of computed tomography reticulation indexes were digitally calculated from trans-axial images using absorptive values, and the computed tomography reticulation indexes compared with clinical scales and measurements. Of 326 patients evaluated by at least one of lymphoscintigraphy, bio-electrical impedance, and computed tomography, 24 were evaluated by all three. The mean number of computed tomography scans in these patients was 109. Sixteen patients had breast cancer, seven had gynecologic cancers, and one had primary lymphedema. Mean computed tomography reticulation index (r = 0.52, p < 0.01) and maximal computed tomography reticulation index (r = 0.45, p < 0.05) were significantly associated with time from initial limb swelling to computed tomography. Mean computed tomography reticulation index (r = 0.86, p < 0.01), minimal computed tomography reticulation index (r = 0.79, p < 0.01), and maximal computed tomography reticulation index (r = 0.68, p < 0.01) were significantly associated with International Society of Lymphedema substage. Minimal computed tomography reticulation index correlated with 1-kHz-based bio-electrical impedance ratio (r = -0.46, p < 0.05) and with standardized proximal limb circumference difference ratio (r = 0.45, p < 0.05) of both limbs. Maximal computed tomography reticulation index had a sensitivity of 0.78, specificity of 0.60, and areas under the curve of 0.66 in detecting lymphoscintigraphic stage IV. The algorithm utilizing three-dimensional computed tomography images of epifascial fibrosis may be used as a marker for lymphedema duration, limb swelling, International Society of Lymphedema substage, and interstitial lymphatic fluids of lymphedema. The current approach shows promise in providing an additional method to assist in characterizing and monitoring lymphedema patients.

Trial registration: ClinicalTrials.gov NCT03523494.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. The images under the proposed image processing procedures.

The original CT image of the affected limb (the first image; the farthest left one in the upper row), six intermediate images and the final image (farthest right in the lower row) are shown. The algorithm eliminated pixels from -50 Hounsfield unit (HU) to 26 HU (image 2), and removed any remained noise using a simple opening filter and manual reoperation (image 3). The outmost skin was removed based on its location (image 4). The remained pattern was processed as one connected region using the closing filter (image 5). Outside of this one connected region, the eliminated skin line was added as one outside thick line (image 6). This merged morphological pattern was converted by the black-white reversion operation. (image 7). The inversed pattern subtracted from the original CT image yielding a donut-shaped region between the skin and the muscle. (image 8). The nonblack pixels of the final image were regarded as epifascial lymph-proliferative fibrosis.

Fig 2. Correlation between the CT reticulation indexes and other clinical measurements.

(A) Maximal CT reticulation index (CTRIMAX) showed significant relationship with time from the onset of lymph edema to CT (LETOCT). (B) Minimal CT reticulation index (CTRIMIN) showed significant relationship with difference in the circumference ratio of an affected and nonaffected limb 5 cm proximal to the anatomical landmarks (cm) (CIRUP). (C) CTRIMIN showed significant relationship with 1 kHz-based bio-electrical impedance ratio (BEI) between an affected and nonaffected limb. (D) Mean CT reticulation index (CTRIMEAN) showed significant relationships with LETOCT. Spearman correlation analysis or Mann-Whitney U test.

Fig 3. Receiver operating characteristics curve analysis of the effects of CT reticulation indexes on the epifascial and subfascial lymphatic system dysfunction.

Maximal value of CT reticulation index can distinguish (A) lymphoscintigraphic stages III and IV from stages I and II and (B) stage IV from stages I, II, and III. AUC. area under the curve; CI, confidence interval.