Quantitative Assessment of Kidney Interstitial Fibrosis Using Optical Spectroscopy Mediated through a Fine Needle Probe

Visual Abstract

Abstract

Key Points

• Intrarenal elastic-scattering spectroscopy at multiple sites accurately quantifies kidney fibrosis in a CKD rat model.

• Machine learning analysis predicted the histologic fibrosis grades with >98% accuracy.

• This method provided a rapid, objective tool for kidney evaluation at the point of care without increasing bleeding complications.

Background

Quantitative assessment of renal fibrosis on biopsy remains a time-consuming step and is subject to interobserver variability, staining inconsistencies, and sampling errors, given the heterogeneous nature of renal fibrosis. An objective, rapid, and point-of-care evaluation method at multiple locations in the kidney is particularly needed for preimplantation donor kidney assessment, where such limitations can lead to the erroneous nonuse of approximately 20% of grafts. We investigated the accuracy of an orthogonal technique—elastic scattering spectroscopy (ESS)—and compared it against the conventional interstitial fibrosis and tubular atrophy (IFTA) scale.

Methods

CKD was induced in rats via an adenine-enriched diet. A customized fine-needle ESS probe was used to acquire spectral data at predetermined intrarenal depths, followed by standard histopathologic scoring. Linear regression was initially used to correlate ESS-derived spectral features with expert-reported IFTA scores. Subsequently, both continuous and categorical random forests (RFs) were implemented to predict fibrosis severity.

Results

RF regression applied to the 300–500 nm spectral range yielded R² values of 0.925, 0.917, and 0.838 at depths of 1.5, 3 mm (renal cortex), and 5 mm (corticomedullary junction and medulla) depths, respectively. Averaging across depths increased the R² to 0.961. Corresponding categorical RF accuracies were 98.14%, 96.74%, and 93.49%, with an averaged overall accuracy of 98.61% following five-fold cross-validation.

Conclusions

Intrarenal ESS provides a robust, depth-resolved, and quantitative estimation of fibrosis that closely parallels histologic IFTA scoring. Multidepth sampling improves predictive accuracy by minimizing local variability. ESS thus represents a promising, objective, rapid, and scalable adjunct for point-of-care kidney tissue evaluation at multiple points within the kidney without increasing risk of bleeding.

Introduction

Renal fibrosis is the final common pathway of CKD and the principal structural determinant of irreversible renal functional decline.^1,2 It is characterized by interstitial fibrosis and tubular atrophy (IFTA), a dominant feature of chronic interstitial nephritis and calcineurin inhibitor nephrotoxicity, and a central determinant of long-term renal prognosis.^1,3,4 The extent of renal fibrosis has substantial prognostic and therapeutic implications. Across diverse kidney diseases—including kidney transplantation, diabetic kidney disease, lupus nephritis, FSGS, and other primary glomerulopathies—IFTA severity consistently predicts loss of kidney function, independent of glomerular pathology, and can directly inform the therapeutic strategies.^5–12

Kidney biopsy with histologic assessment using Masson trichrome staining remains the gold standard for estimating renal fibrosis. However, this approach is constrained by several intrinsic limitations. Tissue processing typically requires ≥48 hours and is subject to variability in staining protocols and interobserver differences in semiquantitative IFTA scoring. More fundamentally, biopsy sampling is inherently limited in spatial representation. A standard 16-gauge or 18-gauge core biopsy samples approximately 31 mm³ of tissue—<0.02% of the average adult human kidney volume (approximately 154,000 mm³). The fibrosis estimate derived from this minuscule sample is extrapolated to the entire organ, despite the well-established spatial heterogeneity of renal fibrosis. Consequently, biopsy-derived fibrosis assessment is prone to substantial sampling error, with focal overestimation or underestimation of disease burden depending on the site of tissue acquisition. Because clinical decisions are often based on a single biopsy, opportunities for spatial averaging or reassessment are limited.

These limitations are particularly consequential in kidney transplantation. Donor kidneys typically undergo a single preimplantation biopsy evaluated by frozen section, and the degree of IFTA and glomerulosclerosis heavily influences accept-reject decisions. Up to 20% of procured donor kidneys are discarded, in significant part because of erroneous, nonrepresentative, or artifact-prone biopsy findings.^13–18 Together, these constraints highlight an unmet need for a rapid, objective, and spatially representative method for renal fibrosis.

Elastic scattering spectroscopy (ESS) is a point-of-care optical spectroscopy technique that provides rapid (<1 minute), quantitative assessment of tissue microstructure. Unlike imaging modalities, ESS performs localized volumetric spectroscopic interrogation of a three-dimensional tissue (approximately 0.2 mm³) surrounding the probe tip. The resulting signal represents a biophysical signature of the interrogated region, reflecting features such as cellular density, nuclear morphology, and extracellular matrix organization. ESS leverages wavelength-dependent photon scattering to characterize tissue microarchitecture without the need for staining, fixation, or prolonged processing.^19–27 Importantly, ESS can be performed minimally invasively at multiple depths and locations, enabling broader spatial sampling of this heterogeneous disease process.

We previously demonstrated a strong correlation between ESS spectral features and renal fibrosis in ex vivo exsanguinated kidney cores.²⁸ However, whether ESS could provide accurate, real-time in vivo fibrosis assessment remained unknown. In this study, we evaluate the performance of in vivo ESS for the quantitative assessment of renal fibrosis using a rat model of CKD. By directly comparing multidepth intrarenal ESS measurements with IFTA scoring and integrating machine learning–based analytical approaches, we test the hypothesis that ESS can provide rapid, accurate, and spatially robust fibrosis assessment. These studies establish the translational feasibility of ESS as a complementary point-of-care tool for kidney evaluation.

Methods

Rat Models and ESS Data Acquisition

The Institutional Animal Care & Use Committee of Boston University, IACUC202300000007, approved the experimental protocol, and all procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The experiment used female albino rats (8–12 weeks) weighing 200–250 g, Strain Code 001 (Charles River Laboratories). The animals were housed under standard laboratory conditions, maintained on a 12-hour light/dark cycle, and provided with food and water ad libitum. Animals were randomly assigned to the experimental groups. CKD was induced by a 0.75% adenine diet, Teklad Rodent Chow (C20070701), for 2 weeks before the procedures and was continued postoperatively. Groups of three rats were harvested at 1-week intervals for 4 weeks (Figure 1A).

Figure 1.

Schematic of in vivo experiment and ESS data acquisition. (A) Schematic timeline depicting study design of rats over 4 weeks. Three animals were assigned per group at each time point (7, 14, 21, and 28 days). The figure demonstrates the biopsy and ESS collection time points for control and CKD rats. (B) Path of light from ESS broadband excitation source through a fiberoptic to ESS probe into the tissue. The light scatters in the tissue and is captured with the return fiberoptic, which is directed to a spectrometer and software for analysis. (C) The depths of assessment of renal tissue with fine needle ESS probe at 1.5, 3, and 5 mm. ESS, elastic scattering spectroscopy.

Rats were anesthetized with 4%–5% isoflurane induction and maintained with 1%–2% Isoflurane. The abdominal cavity was accessed through a midline incision under sterile conditions. The kidneys were gently mobilized from the surrounding areolar tissue using blunt dissection. A surgical loupe with 3.5× magnification and an light-emitting diode light source was used in the dark room to minimize background light, which minimizes background noise for the ESS spectral measurements. The 23-gauge needle probe was inserted into the kidney's medulla at five different loci, at three depths of 1, 3, and 5 mm, starting from the lower pole and proceeding to the upper pole semicircularly. A 20-microsecond light pulse was delivered from a xenon arc lamp through a 150-μm optical fiber, and backscattered light from the tissue was collected by an adjacent 100-μm fiber and conducted to a spectrometer to generate an ESS spectrum.

The measurements were taken quickly, based on depth indicators on the needle, to minimize blood at the fiber probe tip, which would have interfered with the backscattered light data. After each animal measurement, the used needles were removed from the probe and discarded, while the fiber probe was cleaned with isopropyl alcohol for reuse. After the measurements were taken, the animals were euthanized, and the kidneys were harvested.

Histologic Assessment

Hematoxylin and Eosin (H&E) and Masson Trichrome stains were performed on the slides in the Histopathology Core at Boston University Medical Center. The slides were evaluated by a nephropathologist.

Tubulointerstitial Assessment

Tubular atrophy, dilatation, casts, interstitial inflammation, and fibrosis as parameters of interstitial damage were determined using a semiquantitative IFTA scoring method on periodic acid-Schiff–stained sections. IFTA grade 0 indicates no changes; grade 1, fibrotic lesions affecting <25% visualized area; grade 2, lesions affecting 25%–50%; grade 3, affecting 50%–75%; and grade 4, lesions involving more than 75% of the field.²⁸

ESS

ESS provides spectra sensitive to cellular and subcellular architecture changes associated with tissue pathologies. ESS collects backscattered light from tissue near the probe tip over a broad wavelength range (320–900 nm in our current system), using a fiberoptic geometry that enhances sensitivity to cellular and tissue architecture and extra cellular matrix micromorphology (Figure 1B). ESS is a form of subdiffuse reflectance spectroscopy, specifically designed for small source-detector separations (approximately 150 µm in this case). The diffusion approximation to the transport equation is invalid at such small separations, and sensitivity to the scattering phase function is enhanced.²⁹ With this optical geometry, ESS is sensitive to the absorption spectra of strong chromophores (e.g., oxy-/deoxy-hemoglobin) and, importantly, the scattering properties of subcellular structures and the extracellular matrix.

The probe consists of two elements: optical fibers and an outer casing. The outer casing comprises four parts: a syringe body, a plunger, a cap, and a luer lock tip (Supplemental Figure 1). The design allows the user to extend the optical fiber probe slightly beyond the needle tip with only one hand. The 23-gauge needle, from which the fiber probe can extend and retract, is attached to the casing with a standard luer lock. This size causes minimal damage to the kidneys. The probe can be inserted interstitially to arbitrary depths (in this study: 1.5, 3, and 5 mm; Figure 1C), reaching different anatomic sections of the kidney (cortex, interface, and medulla, respectively). The intermediate depth of 3 mm is presumed to probe the interface between the cortex and the medulla.

Analysis of ESS Data and Statistics

Before recording spectra from the tissue, a reference spectrum from a spectrally flat diffuse reflection standard (Spectralon, Labsphere, Inc.) is taken, yielding a reference spectrum, I (λ)_ref. The diffuse reflectance of this reference is >98% for the full spectral range of the ESS measurement. The reference spectrum is used to correct for the system response, including any spectral variations in the light source, spectrometer, fiber transmission, and fiber coupling. Each ESS spectrum in this study was recorded as the ratio of the tissue spectrum to the reference spectrum, providing spectra independent of the system’s spectral response. The optical system, including the probe, was calibrated at the time of the measurement, followed by the tissue measurements. Spectral measurements were triggered using either a foot pedal or the keyboard on the integrated computer. Although the fast optical gating results in minimal background light effects, the instrumentation compensates for any background light and detector offset or dark current by taking an initial measurement without firing the xenon lamp $\left({I\left[\lambda \right]}_{tissue\hspace{0.17em}background}\hspace{0.17em}\text{or}\hspace{0.17em}{I\left[\lambda \right]}_{ref\hspace{0.17em}background}\right)$, followed immediately by a subsequent reference or tissue measurement with the pulsed lamp triggered. The calculated ESS spectra result from subtracting the background spectrum from the tissue or reference spectra. Thus, calibrated spectra are calculated according to:

$$I\left(\lambda \right)=\frac{{I\left(\lambda \right)}_{tissue}-{I\left(\lambda \right)}_{tissue\hspace{0.17em}background}}{{I\left(\lambda \right)}_{ref}-{I\left(\lambda \right)}_{ref\hspace{0.17em}background}}$$

Results

Adenine-Induced CKD in Rats

We used an adenine-induced rat model of CKD.^30–33 Adenine transforms to 2,8-dihydroxyadenine in the liver and then is deposited as crystals of 2,8-dihydroxyadenine in the renal tubules due to concentrated urine in rodents. A 28-day laboratory protocol (Figure 1A) was used, where the rats were given a 0.75% adenine diet and kidneys were harvested at specific time points: 7, 14, 21, and 28 days. This strategy enabled longitudinal follow-up of renal fibrosis estimation in the kidneys of rats undergoing CKD. The perfused rat kidneys underwent ESS measurement using the newly developed probe and prespecified locations (see Methods section). After each weekly set of ESS measurements on the exposed kidneys, the kidneys were harvested, fixed, and stained with H&E and Masson-trichrome application at the site or one closer to the probe insertion (Figure 2, B and C).

Figure 2.

Experimental timeline and histopathologic characterization of renal fibrosis progression in the AD rat model. (A) Representative histologic image of a kidney section showing the measurement depths used for optical spectroscopy (1.5, 3, and 5 mm). Adjacent panels highlight key pathologic features at later fibrosis stages. (B) H&E staining of kidney sections from CTL and AD rats at 7, 14, and 21 days (10× magnification). (C) Masson Trichrome staining of kidney sections from control and AD rats at 7, 14, and 21 days (10× magnification, upper panels; 40× magnification, lower panels). AD, adenine diet; CTL, control diet; H&E, Hematoxylin and eosin.

Evaluation of Renal Histology

Adenine-induced damage resulted in profound changes in the kidneys of rats characterized by the loss of the brush border, tubular atrophy, and dilation. Interstitial inflammatory infiltrates were found throughout the kidney (Figure 2A, top right panel). The tubular basement membrane was thickened along with interstitial fibrosis. On several occasions, adenine crystals were found to be deposited within the kidney tissue, appearing as brown crystals around the peritubular area (Figure 2A, right lower panel).

Figure 2, B and C, shows the longitudinal changes in the renal interstitium. H&E stain showed the presence of adenine crystals, tubular atrophy, and interstitial fibrosis, which increased from day 7 to day 21 (Figure 2B). Masson-trichrome stains fibrosis blue, as shown in Figure 2C. Control kidneys showed minimal trichrome stain in the periglomerular and peritubular regions. However, rats exposed to a 21-day adenine diet showed a striking increase in blue stain around the peritubular and cortical areas (Figure 2C).

The amount of IFTA was measured at the three specified depths by an expert nephropathologist (Mostafa Belghasem).²⁸ Three areas were evaluated, as measured from the surface of the kidneys: 1.5, 3, and 5 mm. At the end of the study, rats on a normal diet exhibited minimal fibrosis over the 4 weeks (ND1–ND4; Figure 3A). Rats on an adenine diet showed progressively increasing IFTA at all depths from week 1 to week 4. The highest IFTA was noted within 5-mm depth in rats after 4 weeks of the adenine diet. Some degree of fibrosis was reported in the first week at depths of 3 mm and 5 mm, and this trend persisted into the third week, where the 5-mm depth showed slightly higher fibrosis compared with the other depths. At week 4, however, all the depths showed comparable amounts of IFTA.

Figure 3.

Quantitative assessment of IFTA scores across different needle depths and time points in normal and AD-induced renal fibrosis. (A) Comparison of IFTA scores at needle insertion depths of 1.5 mm (blue circles), 3 mm (red squares), and 5 mm (purple triangles) across ND and AD groups at various stages of fibrosis progression (ND1–ND4 and AD1–AD4 indicate sequential sample time points or stages. (B–E) Temporal evaluation of IFTA scores in ND (blue circles) and AD (red squares) groups at the first (B), second (C), third (D), and fourth (E) weeks postinduction. Measurements were recorded at 1.5, 3, and 5 mm needle depths. AD groups demonstrate progressively elevated IFTA scores compared with ND controls, with higher scores correlated with deeper needle insertions and advancing fibrosis over time. Data points represent individual measurements. IFTA, interstitial fibrosis and tubular atrophy; ND, normal diet.

IFTA Histopathologic Results Compared with ESS

The spectral data from each rat were obtained and labeled with the rat number, depth of interrogation, and CKD/non–CKD-induced qualifiers. There were 1075 spectral samples per depth, resulting in 3225 total spectra. For each spectrum, a five-point moving average filter was used for smoothing (over the 2048 pixels scanning the spectrum), and all spectra were normalized to a value of 1 at the wavelength 650 nm to permit comparison of spectral shapes, independent of overall amplitude.²⁸ Figure 4A was generated by plotting the averaged readings of all depths for each week to help visualize the progression of spectral change due to CKD induction. The traces of each CKD rat in Figure 4A were averaged together for each week to produce Figure 4B. On inspection of Figure 4, the most significant spectral alterations included an increase in slope between 312 and 375 nm, an increase in the area under the curve between 530 and 580 nm, and a decrease in relative depth of the Q-band (one of the dominant spectral features due to optical absorption by hemoglobin) between 500 and 600 nm from weeks 1 to 4.²⁸ The spectral changes for the non-CKD rats were minimal from weeks 1 to 4 as expected, as presented in Figure 4C. We then sought to correlate changes in ESS spectral features with the IFTA scores reviewed above.

Figure 4.

Averaged spectra per week for non-CKD and CKD rats. The overall spectral changes that take place from weeks 1 to 4 for CKD and non-CKD rats at all depths. (A) Averaged spectra obtained from each CKD rat. The traces represent averaged measurements at all depths for each CKD rat weekly. For clarity, individual control rat spectra were omitted and replaced with the weekly averaged spectra. All spectra were also normalized to 650 nm after performing a five-point moving filter smoothing. (B) Individual traces for each CKD rat in (A) were averaged together each week. A five-point moving average filter was applied for smoothing, and each spectrum was normalized to 650 nm to permit comparison of spectral shapes and relative amplitudes. (C) Weekly averaged ESS spectra for non-CKD rats only. For each non-CKD rat, depth-averaged spectra were computed and then averaged per week. (D) Comparison of week 1 non-CKD spectra with representative CKD spectra from weeks 2 to 4.

Given observable spectral differences between weeks and a significant weekly increase in IFTA scores, we sought a method to quantify these differences and relate spectral changes to IFTA scores in a quantitative correlation. We used linear and logistic regressions initially to assess the spectral features, including various slopes, the area under the curve, and relative Q-band depth, individually and together against IFTA score, but the end results were only weakly associated, with R² values under 0.7. This was not surprising, given the complexity of the spectra, which invoked changes in both the scattering and absorption coefficients. Therefore, we decided to leverage a machine-learning method based on random forests (RFs), due to their ability to effectively handle high-dimensional data and their inherent robustness against overfitting, especially in multiclass settings and in the presence of nonlinearity.^34,35

The RF model used the 300–500 nm range, which was qualitatively selected and informed by a previous ESS renal fibrosis study demonstrating strong collagen scattering in the 350–450 nm range. Other combinations of ranges beyond 500 nm were also considered but offered lower accuracies due to lower signal-to-noise ratio.²⁸ The model using the 300–500 nm range leveraged the random sampling and bootstrap aggregation features of the RF algorithm to select significant spectral features and build a decision tree. IFTA was first assessed as a continuous variable to facilitate quantification of CKD progression and then categorically to simplify predictions into more clinically relevant stages. The IFTA categories were determined to be <0.6, 0.6–1.2, 1.2–1.8, and >1.8 for mild, moderate, moderate-severe, and severe fibrosis. The dataset was partitioned using 80% to train the model and 20% to test the model. The model performance statistics presented are based on the 20% testing dataset outputs. After assessing each model, a five-fold cross-validation was also used to assess the reproducibility and accuracy of the models. Table 1 presents the results of the cross-validation, which display accuracies similar to those of their respective models for all depths. Additional model metrics are presented in Supplemental Tables 1 and 2.

Table 1.

Model performance summary

Depth (mm)	Full Spectrum Regression R2 (MSE)	Full Spectrum Model % Accuracy (% Accuracy in Five-Fold Cross-Validation)
1.5	0.925 (0.0331)	98.14 (97.02)
3	0.917 (0.0294)	96.74 (96.00)
5	0.839 (0.0806)	93.49 (94.98)
AVG	0.961 (0.0181)	98.61 (97.21)

Summary of depth-wise performance statistics for random forest regression and the model from Figure 5. The R² regression models and mean squared error are presented. The percent and average percent accuracy from a five-fold cross-validation are presented for the full-spectrum continuous model. AVG, the average of the measurements at depths 1.5, 3, and 5 mm; MSE, mean squared error.

In the continuous IFTA model, the 1.5-mm depth exhibited an mean squared error (MSE) of 0.033 and an R² of 0.925. The scatter plot in Figure 5A helps to confirm this accuracy due to increased clustering around the red perfect prediction line with low point spread. The 3-mm depth also performed with an MSE of 0.029 and an R² value of 0.917. The 5-mm depth exhibited a decline of model performance, evident by the lower R² of 0.839 and higher MSE of 0.081. Figure 5E further supports this observation, as corroborated by the increased point dispersion seen with decreased clustering around the perfect prediction line.

Figure 5.

RF model of 300–500 nm spectral range. The results were demonstrated using the 300–500 nm spectrum. Left column: actual versus predicted continuous IFTA scores, depth-wise and averaged from a RF regression analysis. The red line is the perfect prediction line where predicted and actual IFTA scores are identical. Right column confusion matrices detailing the categorical IFTA-based RF model assessment output. The horizontal axis represents the predicted class, and the vertical axis represents the actual class. The blue diagonals of the matrix represent the data in which the spectral classification was correctly predicted. Data points below this diagonal represent cases where the algorithm underestimated the IFTA score, and points above the diagonal represent overestimations. RF, random forests.

The 300–500 nm range RF categorical model displayed the higher performance statistics. The 1.5-mm depth yielded a 98.1% accuracy with a macro F1 score of 0.958. The confusion matrix in Figure 5B also supports this in showing minimal misclassifications. The 3-mm depth demonstrated accuracy of 96.7% and slightly increased misclassifications (Figure 5D) as compared with the 1.5-mm depth. The 5-mm depth yielded an accuracy of 93.5%, macro F1 score of 0.872, and more misclassifications, as seen in Figure 5F, which is a decline in performance compared with the shallower depths.

By inspection of the graph in Figure 6, it appears that the IFTA scores do not significantly differ between depths. Based on the results provided by the RF model, it also appears that spectra correlate well to their respective IFTA scores. Therefore, the depths were averaged for each respective spectra, and the RF model steps were repeated (including the same region of analysis and RF model parameters). The RF continuous model for the averaged depths yielded a R² value of 0.961 with MSE of 0.018. The RF categorical model exhibited an accuracy of 98.6% with macro precision of 0.995, macro recall of 0.970, and macro F1 score of 0.981. Thus, the continuous and categorical RF models for averaged depths demonstrated higher accuracies compared with the depth-specific models, especially compared with the lower depths.

Figure 6.

IFTA progression from weeks 1 to 4 based on histopathology results. The progression of IFTA scores as recorded by expert pathologist from weeks 1 to 4 at each depth. The x-axis represents the week of the experiment and whether the rats were non-CKD (1c, 2c, 3c, 4c) or CKD induced (1, 2, 3, 4). The y-axis represents the IFTA score. SD error bars are also presented.

Discussion

We demonstrate the feasibility of an orthogonal approach for quantitative renal fibrosis assessment using an objective, rapid, portable, and scalable ESS platform. Using a fine-needle intrarenal probe in a rat model of CKD, ESS measurements showed a strong concordance with the gold-standard Masson trichrome–based histologic assessment of fibrosis, achieving a final predictive accuracy of 98.61%. Both regression-based (continuous) and categorical RF machine-learning models yielded accurate and highly precise predictions across all sampled depths. The highest single-depth accuracy was observed at 1.5 mm, corresponding to the renal cortex, consistent with the known predilection for early fibrotic remodeling in this compartment in the adenine-induced CKD model. By contrast, deeper measurements at 5 mm, sampling the corticomedullary and medullary regions, demonstrated a modest decline in predictive accuracy, likely reflecting lower overall fibrotic burden and increased spatial heterogeneity in these regions. Importantly, this depth-dependent attenuation reflects the underlying biology of renal fibrosis rather than a technical limitation of ESS, which interrogates a consistent localized tissue volume at all depths. Notably, averaging ESS measurements across all depths produced the most robust predictive performance, mitigating the impact of regional variability and yielding an integrated representation of heterogeneous intrarenal fibrosis.

From a sampling geometry standpoint, a standard 16-gauge biopsy core has an in-plane diameter of approximately 1–2 mm and a length of approximately 1 cm, from which a two-dimensional histologic section is assumed to be representative of the entire sampled volume. By contrast, ESS interrogates optical scattering from a localized three-dimensional microenvironment at each measurement depth, with submillimeter lateral resolution. Because ESS does not require tissue removal, multiple depth-resolved measurements can be obtained during a single fine-needle pass, or through several low-trauma passes, allowing broader lateral and axial sampling of the kidney parenchyma.

Although perfusion measurements were not incorporated in this study, ESS is also capable of extracting hemodynamic information, enabling assessment of the renal microvasculature at multiple intrarenal sites without the need for biopsy.²⁷ By permitting repeated, spatially distributed intrarenal measurements, ESS offers a fundamentally different and potentially superior approach to evaluating heterogeneous renal fibrosis compared with single-site core biopsy.

It is also essential to contextualize ESS relative to other advanced optical microscopy and imaging modalities used for fibrosis assessment. Techniques such as second-harmonic generation and fluorescence-lifetime imaging microscopy provide high-resolution, collagen-specific structural information but require an excised biopsy specimen that must be processed ex vivo before imaging.³⁶ As such, these approaches remain subject to the same fundamental sampling constraints as conventional histology. By contrast, ESS is performed in situ, without tissue removal, and provides immediate spectroscopic information rather than architectural images. Whereas histology and second-harmonic generation–based approaches generate a two-dimensional snapshot from a single cylindrical core, ESS interrogates a three-dimensional, localized tissue volume surrounding the probe tip. Monte Carlo simulations in this configuration estimate an interrogated volume of approximately 0.2 mm³. This volumetric spectroscopic sampling captures the intrinsic tissue-scattering properties and enables serial, multisite interrogation of heterogeneous renal parenchyma. This distinction is particularly critical given that conventional biopsy samples only account for approximately 0.02% of the total human kidney volume and require irreversible tissue extraction. Cross-sectional imaging modalities such as magnetic resonance imaging can estimate global fibrosis three-dimensionally but cannot simultaneously assess perfusion in specific local areas of the kidney.³⁷ Ultrasound-based techniques such as B-mode sonography and elastography are limited by subjectivity in operator data acquisition and a modest ability to resolve early or heterogeneously distributed fibrosis.

This study extends our previous observations to a rat model of CKD, with direct translational relevance to cadaveric donor kidney evaluation. Given the persistently high nonuse rate of deceased donor kidneys in the United States,³⁸ ESS has the potential to serve as a valuable adjunct to conventional pathologic assessment. The rapid, spatially distributed interrogation enabled by ESS uniquely positions this technology to transform donor kidney evaluation by providing real-time, reproducible, and comprehensive assessment of renal microarchitecture and macroarchitecture at the point of care. Beyond kidney transplantation, this modular diagnostic framework establishes the foundation for broader application across other solid-organ transplants and the development of predictive models of graft quality and post-transplant outcomes. Because ESS is independent of histologic staining, it eliminates interobserver subjectivity and staining-related variability inherent to traditional pathology. In addition, ESS can quantify both oxygenated and deoxygenated hemoglobin during surface interrogation. Although this perfusion-related parameter may be less applicable in deceased kidneys undergoing machine perfusion, it remains highly relevant for interrogating renal vascularity, a critical determinant of graft survival. While the current application of ESS is focused on fibrosis detection, the underlying artificial intelligence and machine learning framework is adaptable and can be trained to recognize additional known and previously underappreciated tissue features, further expanding its diagnostic potential.

This study has limitations. Although the adenine model recapitulates key features of tubulointerstitial fibrosis, future studies incorporating additional CKD models will be required to establish the broader generalizability of this approach.³⁹ Only female rats were studied. Although sex-specific effects could not be directly assessed in this proof-of-principle investigation, prior work demonstrates that adenine-induced CKD produces comparable histopathologic fibrosis in both sexes.⁴⁰ We anticipate that optical scattering behavior would be similar in male rats. Finally, some degree of tissue shrinkage is unavoidable due to formalin fixation. ESS captures the scattering properties before excision. Since the shrinking is mainly uniform, the local anatomic relationships at the measurement depth are largely preserved. Regardless, this fixation-related distortion remains an inherent limitation.

In conclusion, this study establishes ESS as a powerful, objective, and minimally invasive platform for quantitative assessment of renal fibrosis in both native and transplant kidneys. By demonstrating strong concordance with gold-standard histopathology and exceptionally high predictive accuracy across multiple intrarenal depths, ESS overcomes key limitations of conventional biopsy, including sampling error, processing delays, and interpretive variability. These unique features position ESS as a transformative leap in organ evaluation, with the potential to redefine donor selection, substantially reduce unnecessary organ discard, and meaningfully expand the pool of transplantable kidneys, thereby delivering a durable and immediate impact on transplantation outcomes. This relevance is further strengthened by the recent US Food and Drug Administration clearance of ESS, underscoring the clinical maturity and regulatory readiness of this technology.⁴¹ With the potential to reduce erroneous organ nonuse and expand access to transplantation, this work provides a compelling foundation for the clinical translation of ESS as a next-generation adjunct to the on-call transplant pathologist.

Supplementary Material

kidney360-7-1463-s001.pdf ^{(1.4MB, pdf)}

kidney360-7-1463-s002.pdf ^{(350.3KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/KN9/B594.

Author Contributions

Conceptualization: Irving J. Bigio, Vipul C. Chitalia, Jean Francis.

Data curation: Mostafa Belghasem, Vipul C. Chitalia, Ji Young Chung, Meghan Howard, Saran Lotfollahzadeh, Emily Oros.

Formal analysis: Ousama A’amar, Mostafa Belghasem, Irving J. Bigio, Parth Jalihal, Saran Lotfollahzadeh, Eladio Rodriguez-Diaz.

Funding acquisition: Vipul C. Chitalia.

Investigation: Vipul C. Chitalia, Saran Lotfollahzadeh.

Methodology: Vipul C. Chitalia, Saran Lotfollahzadeh.

Project administration: Vipul C. Chitalia.

Resources: Vipul C. Chitalia.

Software: Vipul C. Chitalia.

Supervision: Vipul C. Chitalia.

Visualization: Ousama A'amar, Mostafa Belghasem, Irving J. Bigio, Vipul C. Chitalia, Parth Jalihal.

Writing – original draft: Ousama A'amar, Mostafa Belghasem, Irving J. Bigio, Vipul C. Chitalia, Jean Francis, Parth Jalihal, Aman Kumar, Saran Lotfollahzadeh, Sayeed Malik, Eladio Rodriguez-Diaz.

Writing – review & editing: Ousama A'amar, Mostafa Belghasem, Irving J. Bigio, Vipul C. Chitalia, Jean Francis, Parth Jalihal, Aman Kumar, Saran Lotfollahzadeh, Sayeed Malik, Eladio Rodriguez-Diaz.

Funding

V.C. Chitalia: NHLBI Division of Intramural Research (R01HL166608). S. Lotfollahzadeh: NHLBI MultiDisciplinary Training in Cardiovascular (T32HL125232), Boston University Chobanian and Avedisian School of Medicine (Fibrosis Affinity Research Collaborative).

Declarative Statements

All animal experiments were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals or an equivalent standard that meets or exceeds the ethical and welfare requirements outlined in the NIH Guide. All protocols were approved by the appropriate institutional animal care and use committee.

Data Availability Statements

Original data generated for the study will be made available upon reasonable request to the corresponding author. Data Type: Aggregated Data; Image Data. Reason for Restricted Access: The spectral files and histology reports reported and analyzed in this report are available to academic researchers upon request to the corresponding author.

Supplemental Material

This article contains supplemental material online, published as provided by the authors, at http://links.lww.com/KN9/B595.

Supplemental Figure 1. Design of the ESS probe.

Supplemental Table 1. Expanded RF regression metrics for all depths.

Supplemental Table 2. Expanded RF categorical model metrics for all depths.