Regional micro-computed tomography analysis enables longitudinal detection of compensatory ventilation effects in a mouse model of pulmonary fibrosis

Abstract

Accurate lung function assessment is essential in preclinical idiopathic pulmonary fibrosis (IPF) research, yet conventional whole-lung endpoints overlook spatial heterogeneity and compensatory mechanisms. We present a micro-computed tomography (µCT)–based pipeline for longitudinal, regional quantification of structural and functional changes in a murine bleomycin (BLM)–induced fibrosis model to improve evaluation of disease progression and therapeutic effects. C57BL/6 mice received triple oropharyngeal administrations of saline or BLM. Respiratory-gated µCT scans were acquired in free-breathing at baseline and days 7, 14, and 21. A deep-learning algorithm segmented the lungs into left and right lobes, further subdivided into apical and caudal regions using airway landmarks. Regional volumes, aeration compartments, and ventilation maps were extracted to assess structural and functional alterations. Saline-treated mice showed stable metrics over time, with minimal interanimal variability, demonstrating the robustness and physiological consistency of our segmentation procedure. BLM-treated animals exhibited early and heterogeneous fibrotic changes, with the apical regions appearing as most affected. The caudal-right region displayed a compensatory functional increase of respiratory parameters derived from multivolume µCT. µCT-based regional analysis can detect localized dysfunction and compensatory effects not captured by whole-lung measurements. This strategy is applicable to different models of heterogeneous lung disease and may contribute to reduce the translational gap in IPF research.

Clinical Relevance.

Whole-lung readouts, while useful, risk overlooking the spatial heterogeneity and compensatory mechanisms that characterize fibrotic remodeling. By integrating regional structural and functional biomarkers, our pipeline enhances sensitivity to detect early changes and treatment effects in preclinical models. This approach provides a more nuanced understanding of disease dynamics and may improve the predictive value of animal studies for clinical translation.

Introduction

Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive lung disease characterized by excessive extracellular matrix deposition around alveoli, eventually leading to impaired gas exchange and respiratory failure^1,2 With a median survival of only 3-5 years ­following diagnosis, prognosis remains poor.³ The US Food and Drug Administration–approved antifibrotic drugs pirfenidone, nintedanib, and nerandomilast slow but do not halt disease progression, and lung transplantation remains the only curative option.^4-6

Despite major efforts to develop effective therapies, few candidates progress beyond phase 2 or 3 trials,^7,8 reflecting both the biological complexity of IPF and weaknesses in the translational pipeline. A major challenge lies in the mismatch between preclinical and clinical efficacy evaluations: While human trials rely on functional endpoints such as forced vital capacity (FVC) over 6-12 months, preclinical studies often focus on histopathological and biochemical markers,^9,10 limiting translational validity and predictive power.

In preclinical research, pulmonary fibrosis (PF) is most induced with bleomycin (BLM), which mimics key hallmarks of human IPF, such as inflammation and fibrotic remodeling.^11-13 However, BLM-induced PF is regionally heterogeneous, with variable collagen deposition across lung regions.¹⁴ These spatial differences can trigger compensatory ventilatory mechanisms,^10,15,16 meaning that less-injured regions contribute to a larger portion of the overall ventilation to compensate for impaired areas. This redistribution can potentially confound global functional readouts, such as murine FVC, and partly mask early functional deficits, especially when assessed invasively (eg, FlexiVent). To bridge this gap, imaging tools are increasingly implemented in both clinical and preclinical research. Clinically, high-resolution computed tomography (CT) is central to IPF diagnosis and management, detecting hallmark features such as honeycombing and reticulation. Quantitative high-resolution CT, especially when coupled with artificial intelligence, is increasingly adopted as a secondary endpoint in clinical trials to assess structural progression,^17,18 although its broader use is still limited by variability in acquisition and reconstruction protocols.^19,20 In preclinical settings, micro-CT (µCT) is increasingly employed to investigate lung disease in small animals and has demonstrated its potential for longitudinal studies, providing quantitative data for tracking disease progression and therapy response.^15,21,22 Yet, most µCT-based studies rely on global metrics (eg, total lung volume or hyperdense volume) that overlook regional disease heterogeneity, a hallmark of PF.

To address this limitation, recent studies explored regional µCT analysis. Lai et al. introduced standardized volumes of interest around bronchial bifurcations, providing reproducible regional sampling but restricted to small subvolumes and without functional readouts.²³ Deep-learning pipelines enable automated whole-lung segmentation and pathology quantification, yet still operate globally and do not delineate anatomically distinct subregions.²⁴ A reproducible left-right anatomical subdivision has been proposed for longitudinal and multicenter studies,¹⁵ though it may miss functional compensation by individual lobes. Apical-caudal subdivision may account for such intralung variations, but its arbitrary thresholds limit longitudinal robustness and cross-study comparability.²⁵ The lack of a standardized, anatomically meaningful framework for consistent lung subregion segmentation remains a major obstacle to the translational applicability of regional imaging biomarkers.²⁴

To overcome these limitations, we developed a µCT-based method to extract quantitative morphological and functional ­biomarkers from 4 anatomically defined lung subregions (apical-caudal, left-right). Compared to whole-lung biomarkers, this regional approach enhances sensitivity and specificity of disease monitoring, providing a refined, region-graded picture of disease progression and has the potential to improve the assessment of therapy response in future studies, thus strengthening the translational relevance of preclinical trials.

Materials and methods

Experimental animals and ethics

Male C57Bl/6 mice (7-8 weeks; Envigo, Italy) were housed under standard conditions with food and water ad libitum, and acclimated for 7-10 days before experimentation. Welfare was monitored daily by trained staff using a Visual Analogue Scale (VAS, 0-10); humane endpoints included dyspnea, ≥20% body weight loss, or a VAS score ≥6.

All procedures complied with Chiesi Farmaceutici S.p.A.’s intramural animal welfare practices for animal research, European Directive 2010/63/EU, Italian D.Lgs 26/2014, the Guide for the Care and Use of Laboratory Animals (2011), and the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. Experiments were approved by the Italian Ministry of Health (protocol number 809/2020-PR) and the institutional Animal Welfare Body and were conducted at an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)–accredited facility.

Experimental design

Twenty-two mice were enrolled. PF was induced in 11 animals by triple (days 0, 2, 4) oropharyngeal aspiration of 6 µg of BLM (corresponding to 0.25 mg/kg) hydrochloride (Baxter) in 50 µL saline under 2.5% isoflurane anesthesia, as previously described.^13,16 Controls (n = 11) received saline alone.

µCT acquisition protocol

Mice were anesthetized with 2% isoflurane and scanned using a Quantum GX µCT scanner (PerkinElmer, Waltham, MA, USA) at 90 kV and 88 µA in high-speed mode with respiratory gating. Animals were positioned supine, and the thoracic region was centered over the diaphragm. Projections were collected in list mode during a continuous 360° gantry rotation (4 minutes, 14 688 projections, 16.33 ms/frame). End-inspiratory and end-expiratory datasets were retrospectively sorted (∼900 projections each) and reconstructed with Graphics Processing Unit (GPU)-based filtered back projection (Ram-Lak filter, 512³ matrix, 50 µm isotropic voxels). System calibration with standard phantoms was performed monthly for noise, uniformity, low contrast, and resolution.²⁶

Conversion from grey levels to Hounsfield units

Grey level-to-Hounsfield unit (HU) conversion was performed using a custom air-water phantom²⁷ scanned at the beginning and end of each imaging session to ensure system stability (90 kV, 88 µA, 8 seconds, no gating). HU values were obtained through a linear fit between 2 reference points: (−1000 HU, mean air grey level) and (0 HU, mean water grey level), with means averaged from respective regions in the phantom images.

µCT segmentation protocol

Lung segmentation was performed automatically using a U-Net convolutional neural network (CNN), previously developed for µCT-based densitometry in murine models of PF.^15,28 For both end-inspiratory and end-expiratory phases, the model segmented whole lungs and its left/right portions. A multiresolution strategy combined low-resolution coarse mapping with a high-resolution 2.5D CNN for fine segmentation; outputs were merged by voting, achieving a Dice coefficient of approximately 0.97 in validation.²⁸ Airways were semi-automatically segmented (AnalyzeDirect, Overland Park, KS, USA).

µCT-guided regional analysis

Automatic left/right segmentation was further refined into apical and caudal regions, yielding 4 subregions: apical-left (Api L), caudal-left (Cau L), apical-right (Api R), and caudal-right (Cau R) (Figure 1).

Figure 1.

Micro-computed tomography–guided regional lung segmentation and representative quantitative outputs. (A) Manual identification of anatomical landmarks used to define lung subregions viewed in the coronal plane of a representative bleomycin (BLM) animal: Red spheres indicate the first and second bifurcations of the left main bronchus and the origin of the right accessory lobe. Red dashed lines represent the apico-caudal division thresholds (Threshold Left, ThL and Threshold Right, ThR), defined based on these landmarks; black dashed lines indicate the measured distance between bifurcations on the left bronchus. Airways are represented with a 3D green rendering (B) example of the resulting lung segmentation mask into 4 subregions in the coronal plane: apical-left (Api L), caudal-left (Cau L), apical-right (Api R), and caudal-right (Cau R). (C, D) Voxel-wise histograms of lung attenuation (Hounsfield units [HU]) and specific gas volume change (DSVg, mL/g) for each subregion in a representative BLM case at day 21, illustrating regional differences in lung structure and ventilation.

Subdivision was based on manual identification of airway landmarks: the first and second bifurcations of the left main bronchus, and the origin of the accessory lobe on the right (Figure 1A, red spheres). To ensure adequate representation of the Api L region, the midpoint (∼1 mm) between left bronchial bifurcations (Figure 1A, black dashed lines) defined the boundary plane (Threshold Left, ThL) (Figure 1A, red dashed line, left); slices above and below ThL were labeled Api L and Cau L, respectively. On the right, the threshold (ThR) corresponded to the bifurcation of the main bronchus and the accessory branch (Figure 1A, red dashed line, right); slices above and below ThR were labeled Api R and Cau R, respectively. Thus, each scan yielded a lung mask subdivided into 4 subregions (Figure 1B). Voxel-wise HU and change in specific gas volume (ΔSVg) distributions within each subregion were visualized as histograms in a representative BLM case on day 21 (Figure 1C, D), illustrating regional differences in lung density and ventilation. These data were used to derive, for each subregion, ­morphological, densitometric, and functional biomarkers, as summarized in Table 1.

Table 1.

Micro-computed tomography–derived biomarkers.

Category	Description	Unit	Formula
Densitometric biomarkers
Inspiratory scan
NINSP	Number of lung voxels at end inspiration	\	Count of lung voxels in insp
VINSP	Lung volume at end inspiration	mm3	$N_{\mathit{INSP}}·\mathrm{voxel}\mathrm{ }\mathrm{size}$
MLAINSP	Mean lung attenuation at end inspiration	HU	$\sum _{i=1}^{\mathrm{ }{N}_{\mathit{INSP}}}{\left(HU\right)}_i/N_{\mathit{INSP}}$
Air (GasP01)	Volume of air at end inspiration	mm3	$\frac{{\mathrm{V}}_{\mathrm{INSP}}·{\mathrm{MLA}}_{\mathrm{INSP}}}{-1000}$
Expiratory scan
NEXP	Number of voxels at end expiration	\	Count of lung voxels in exp
VolEXP	Lung volume at end expiration	mm3	$N_{\hspace{0.17em}\text{EXP}\hspace{0.17em}}·\mathrm{voxel}\mathrm{ }\mathrm{size}$
MLAEXP	Mean lung attenuation at end expiration	HU	$\sum _{i=1}^{\mathrm{ }{N}_{\hspace{0.17em}\text{EXP}\hspace{0.17em}}}{\left(HU\right)}_i/N_{\hspace{0.17em}\text{EXP}\hspace{0.17em}}$
FRC (GasP02)	Functional residual capacity: volume of air at the end of expiration	mm3	$\frac{{\mathrm{V}}_{\mathrm{EXP}}·{\mathrm{MLA}}_{\mathrm{EXP}}}{-1000}$
%Normo-aerated	Normo-aerated lung parenchyma; reflects the extension of no/mild lesions	%	Percent voxels in range (–630 to –307)
%Hypo-aerated	Hypo-aerated lung parenchyma; reflects the extension of moderate lesions	%	Percent voxels in range (–307 to –45)
%Non-aerated	Non-aerated lung parenchyma; reflects the extension of severe lesions	%	Percent voxels in range (–45 to 430)
Functional biomarkers
TV	Tidal volume	mm3	Air-FRC
Median ΔSVg	Median of ΔSVg	mL/g	50th percentile of ΔSVg distribution
%Normal ventilation	Lung parenchyma with normal ventilation	%	ΔSVg ≥ β SVgINSP ≥ ${\alpha }_I$ SVgEXP ≥ ${\alpha }_E$
%Low ventilation	Lung parenchyma with reduced ventilation, not due to fibrosis	%	ΔSVg < β SVgINSP ≥ ${\alpha }_I$ SVgEXP ≥ ${\alpha }_E$
%Non-ventilation	Lung parenchyma with reduced ventilation due to fibrosis	%	SVgINSP < ${\alpha }_I$ SVgEXP < ${\alpha }_E$

Abbreviations: ΔSVg, change in specific gas volume; FRC, functional residual capacity; HU, Hounsfield units; SVg, specific gas volume; TV, tidal volume.

Morphological features included the volume of each lung subregion (mm³), while densitometry provided mean lung attenuation (HU) and aeration compartments²⁷: normo-aerated (–630 to –307 HU), hypo-aerated (–307 to –45 HU), and non-aerated (–45 to +230 HU), reflecting increasing parenchymal involvement.

Functional biomarkers were derived from µCT-derived ventilation maps,²⁵ generated by deformable registration (demons algorithm) of inspiratory to expiratory scan. A Laplacian filter was applied preregistration to enhance anatomical detail and reduce inflation-related intensity changes. Ventilation was quantified voxel-wise as ΔSVg (mL/g) between inspiration and expiration:

$$\mathrm{\Delta }\mathrm{SVg}\mathrm{ }=\mathrm{ }{\mathrm{SVg}}_{\mathrm{INSP}}-\mathrm{ }{\mathrm{SVg}}_{\mathrm{EXP}}$$

with

$$\mathit{SVg}=\frac{1000}{HU+1000}-0·939\mathrm{ }mL/g$$

where 0.939 mL/g corresponds to tissue-specific volume (1/1$.$065 g/mL).

For each subregion, median ΔSVg was extracted as a summary functional metric. Additionally, voxels were classified into ventilation compartments based on control-derived thresholds:

• Non-ventilated (Non-Vent): inspiratory SVg < αI and expiratory SVg < α_E;

• Low ventilation (Low-Vent): ΔSVg < β with inspiratory and expiratory SVg ≥ α_I and α_E, respectively;

• Normal ventilation (Normal-Vent): ΔSVg ≥ β with inspiratory and expiratory SVg ≥ α_I and α_E, respectively.

α_I and α_E are the 5th percentile of inspiratory/expiratory SVg and β the 25th percentile of ΔSVg.

Statistical methods

Statistical analysis was performed using Prism 8 (GraphPad Software, San Diego, CA, USA). Data are expressed as mean ± SEM unless otherwise specified. One- or 2-way ANOVA with Dunnett or Tukey post hoc tests assessed longitudinal changes versus baseline. Normality was verified by the Shapiro–Wilk test and QQ-plot inspection. Statistical significance was set at P < .05.

Results

The experimental timeline, including triple BLM administration and scheduled µCT scans, is outlined in Figure 2A. Representative structural and functional images (Figure 2B-D) illustrate BLM-induced alterations in lung morphology, aeration, and ventilation.

Figure 2.

Experimental timeline and longitudinal qualitative assessment of BLM-induced lung injury in a representative case. (A) Overview of the experimental protocol, including the schedule of triple BLM instillation and subsequent longitudinal µCT acquisitions. (B) Coronal µCT images at baseline and days 7, 14, and 21 post-BLM in a representative animal, illustrating progressive structural changes. Initial uniform aeration is followed by the emergence and expansion of high-density regions, particularly in the Api L subregion (red arrow), with evidence of airway remodeling and fibrotic distortion by day 21. (C) 3D renderings of the lung in P01 and P02 at baseline, colored by aeration and ventilation compartment, respectively. (D) 3D renderings of the lung in P01 and P02 at day 21 colored by aeration and ventilation compartment, respectively. Abbreviations: Api L, apical-left; Api-R, apical-right; BLM, bleomycin; Cau L, caudal-left; Cau R, caudal-right; P01, End-inspiration; P02, End-expiration; μCT, micro-computed tomography.

Progressive fibrotic remodeling is visualized over time in coronal µCT images (Figure 2B). At baseline, lungs display uniform aeration, with well-defined airways and vessels. By day 7, localized high-density areas emerge, especially in Api L (red arrow), with a transient volume increase. By day 14, these regions expand, with signs of architectural distortion, including airway remodeling, predominantly in apical lobes. By day 21, fibrotic remodeling becomes evident, with heterogeneous scarring mainly affecting apical regions.

Three-dimensional reconstructions of aeration and ventilation compartments (Figure 2C, D) confirmed this progression, revealing a shift from uniform normo-aeration (light-green) and normal ventilation (cyan) at baseline to extensive non-aerated (gray) and non-ventilated (magenta) apical regions by day 21.

µCT lung regional analysis in healthy animals

In the control (saline) group, all parameters remained stable across time, documenting the reproducibility and consistency of our segmentation strategy (Figure 3A-D). Whole-lung volumes remained constant (∼600 mm³; Figure 3A), with subregional volumes reflecting known physiological asymmetries (ie, Api L and Cau R being the smallest and the largest, respectively) preserved at all time-points. Minimal error bars indicated low interanimal variability, underscoring protocol robustness. At the densitometric level, normo-aerated tissue accounted for approximately 80% of total lung at end expiration (68%-83% across subregions; Figure 3B), with minimal variability throughout the study. Functionally, %Normal-Vent (Figure 3C) remained constant, with only minor fluctuations in caudal regions by day 21. Similarly, subregional tidal volumes (Figure 3D) were stable and followed the same anatomical pattern as static lung volumes, thus further supporting physiological consistency of regional ventilation in healthy animals.

Figure 3.

Longitudinal regional micro-computed tomography (µCT) analysis in SAL and BLM-treated animals. (A-D) In SAL controls, whole-lung and subregional volumes, aeration, ventilation, and tidal volumes remained stable over time, demonstrating the reproducibility of the segmentation and imaging protocol. (E-H) In BLM-treated mice, µCT biomarkers revealed progressive and heterogeneous alterations across different lung subregions. Data are presented as mean ± SEM. Two-way ANOVA followed by followed by Dunnett post hoc tests for multiple comparisons against the baseline (#, P<.05; ##, P<.01; ###, P<.001). Abbreviations: Api L, apical-left; Api-R, apical-right; BLM, bleomycin; Cau L, caudal-left; Cau R, caudal-right; SAL, saline; TV, tidal volume; Vent, ventilation; V_P02, Volume at End-expiration.

µCT lung regional analysis in BLM-treated animals

In BLM-treated animals, longitudinal µCT revealed progressive, regionally heterogeneous structural and functional alterations (Figure 3E-H). Whole-lung volume transiently increased at day 7 (P < .001; Figure 3E), driven by volume expansion in Cau R and Api R, while Api L decreased significantly from day 14 onward (P = .0052). Notably, Cau R remained enlarged until day 21. Normo-aerated parenchyma (%Normo, Figure 3F) showed a marked and progressive decline in all regions, with the steepest drop between days 7 and 14. Api L showed an early alteration, with normo-aerated volume dropping to approximately 30% by day 7 and further decreasing to approximately 20% by day 14 (P < .001). Api R followed a similar trend, while caudal regions decreased later (from day 14, P < .001).

Ventilation (%Normal-Vent; Figure 3G), declined significantly at the whole-lung level by day 14 (P = .0045). Api L was the earliest and most affected region, with a significant drop already at day 7 (P = .0097), that declined to approximately 19% by day 21. Api R displayed a delayed but progressive decline, with significant reductions at days 14 and 21 (P < .001). Caudal regions were initially preserved but decreased by day 14 (Cau L, P = .0023; Cau R, P = .0107). By day 21, Cau L continued to deteriorate, while signs of partial functional recovery were observed in Cau R. Consistently, tidal volume (TV; Figure 3H) decreased in all subregions except Cau R. The marked apical and Cau L decline was the result of combined structural and functional alterations occurring in these regions, whereas Cau R, despite persistent structural abnormalities, maintained a relatively stable ventilation at early time-points, displayed a trend toward recovery from day 7 onward, and returned to baseline levels by day 21. These divergent behaviors point to a mismatch between structural and functional impairment, possibly due to regional ventilatory compensation in less fibrotic areas.

A comprehensive overview of µCT-derived biomarkers is shown in Figure S1. All parameters are expressed as fold changes relative to baseline except for non-aerated and non-ventilated voxels, which were close to baseline and therefore expressed as raw percentages. The most pronounced deviations occurred in apical subregions, particularly Api L, featuring early and progressive alterations in both structure and function. Structurally, Api L displayed reductions in gas-related parameters (%Gas_P01, %Gas_P02, %Normo, P < .01) by day 7, flanked by increases in %Hypo (P < .01) and %Non-aerated (P < .05). Functionally, this was paralleled by marked declines in TV, median ΔSVg, and %Normal-Vent (all P < .01) with concomitant increases in %Low-Vent and %Non-Vent (P < .05). Except for %Hypo and %Low-Vent, which plateaued, all parameters continued to diverge from baseline through day 21. Api R followed a similar but milder trajectory, with most parameters altered by day 7, and ventilation deficits appearing later. In Cau L, early functional impairment (reduced TV and ΔSVg, P < .001; increased %Low-Vent, P = .0349) preceded structural remodeling, which became evident by day 14 and worsened through day 21, although overall damage remained less severe than in apical regions. Cau R was the most preserved subregion, with a modest increase in %Gas_P02 (P = .0473) at day 7 and transient declines in ventilation (reduced TV, P = .0407; ΔSVg, P < .001; increased %Low-Vent, P = .0504). Ventilation impairment peaked at day 14 (%Normal-Vent, P = .0387; %Non-Vent, P = .00720), followed by functional recovery by day 21, though residual non-aerated and non-ventilated voxels persisted (P < .05), indicating incomplete ­resolution of fibrotic damage.

Discussion

We applied a µCT-based pipeline for longitudinal regional analysis of lung structure and function in both saline- and BLM-treated mice, enabling quantification of spatially resolved injury patterns and ventilatory impairment in a murine model of PF.

In control animals, lung morphology and function remained stable across all time-points with minimal interanimal variability. This confirmed that the µCT acquisition protocol, including anesthesia and the apical/caudal segmentation strategy, caused no appreciable alterations.²⁷ The observed physiological asymmetries, such as reduced volumes and aeration in Api L, were consistent with known murine anatomy, validating the reproducibility and physiological compliance of the proposed subregional framework.

In contrast, BLM administration induced marked, spatially heterogeneous injuries. Apical regions, particularly Api L, were the earliest and most severely affected, displaying increased density and functional decline by day 7. These changes progressed over time, with reductions in TV, normo-aerated, and ventilated compartments, and parallel increases in non-aerated and non-ventilated tissue. This suggests that Api L is particularly sensitive to early fibrotic remodeling and endowed with a limited reparative capacity that closely resembles the irreversible fibrosis seen in human IPF. Api R exhibited a similar but milder trend, suggesting bilateral vulnerability of upper lobes. Distinct temporal dynamics characterized the caudal regions. In Cau L, functional deficits such as reduced ventilation and altered ΔSVg preceded structural changes, suggesting early subclinical dysfunction that may evolve into overt PF. Cau R, instead, displayed the greatest resilience, with transient ventilation decline at day 14 and near-complete functional recovery by day 21, despite persistent fibrotic lesions and reduced aeration. This suggests the activation of compensatory mechanisms rather than spontaneous resolution of PF,¹³ aligning with previous studies showing functional compensation in spared lung regions following injury or lobectomy.²⁹ Apical injury and caudal compensation may reflect bleomycin deposition, mechanical heterogeneity, and perfusion gradients. Similar concepts have been proposed for human IPF, where regional differences in collapse, compliance, and strain influence PF progression.^30,31 These susceptibilities are relevant for therapeutic evaluation, as treatment effects may vary across lung regions with distinct mechanical and biological characteristics, and regional imaging biomarkers may detect localized responses more sensitively than whole-lung endpoints.

This apical-predominant pattern differs from human IPF, yet regional analysis remains essential for interpreting compensatory responses and structure–function relationships.

Regional patterns reveal insights, but histological validation is not feasible. Tissue fixation, embedding, and sectioning alter lung geometry, preventing ex vivo reproduction of the 4 in vivo µCT subregions; only left-right comparisons are reliable. Nevertheless, side-level correlations between µCT biomarkers and histological PF scores have been demonstrated,¹⁵ supporting the biological relevance.

The ability of µCT to detect early regional dysfunctions has important implications for therapeutic evaluation. Early impairment in the apical regions, coupled with compensatory responses in less-affected lobes, underscores the added value of a regionally resolved analysis. Whole-lung endpoints, such as global pulmonary function assessments, may overlook localized improvements or deteriorations, especially at early disease stages. In contrast, regional biomarkers offer enhanced sensitivity to spatially heterogeneous and time-dependent effects, allowing more precise monitoring of therapeutic efficacy and disease progression. This is particularly relevant for antifibrotic treatments, whose effects may vary across lung regions and over time. Such spatial and longitudinal insights are critical for dosage optimization, responders’ identification, and the refinement of therapeutic strategies.

Along this view, µCT imaging should be considered complementary to traditional functional techniques such as forced oscillation or plethysmography, which provide global lung mechanics but lack spatial resolution. While these methods remain valuable for overall respiratory assessment, they may miss subtle, region-specific alterations preceding global dysfunction. Integrating µCT-derived regional biomarkers with standard functional readouts can thus improve accuracy, interpretability, and translational value in preclinical evaluations, supporting more personalized approaches to pulmonary diseases.

Beyond PF, the proposed pipeline holds promise for various pulmonary diseases characterized by regional heterogeneity, including asthma, emphysema, and viral pneumonia. In these contexts, regional µCT can delineate distinct structural and functional patterns, track pathology progression, and identify the most appropriate interventions with spatial precision. Longitudinal tracking of localized changes in aeration and ventilation may also clarify disease mechanisms, such as airway remodeling or alveolar disruption.

While demonstrating high spatial and temporal resolution, some limitations should be acknowledged. The pipeline relies on an automated left/right lung segmentation,^15,28 which, although accurate and reproducible, may not be readily implementable in all laboratories due to computational or technical constraints. Airway segmentation represents a bottleneck as it remains manual or semi-automatic, limiting full automation and scalability. To improve efficiency and reproducibility, future refinements should focus on integrating robust, deep-learning-guided airway segmentation algorithms for high-throughput analysis. Finally, the 4-region division captured major heterogeneities, but finer segmentation could improve sensitivity to early changes. Despite these limitations, our approach captures key features of fibrotic remodeling and regional ventilatory mechanics, often missed by global assessments. Early dysfunction in apical regions, combined with compensatory responses in less-affected lobes, highlights the importance of regionally resolved analysis for understanding disease burden and treatment response. Our results challenge full reliance on whole-lung endpoints and support the integration of imaging-guided, spatially resolved biomarkers in preclinical studies to improve the translational relevance of animal models in IPF research.

Author contributions

M.B., F.P. and F.F.S. conceived and designed the research. E.F. performed in vivo experiments. M.B. and D.B. performed data acquisition. M.B. and F.P. analyzed and/or interpreted data. M.B. was in charge of the artwork and figure assembly. M.B., F.P. and F.F.S. drafted the manuscript. D.B., N.S., E.P. and A.A. edited and revised the manuscript. All authors read and approved of the final manuscript.

Supplementary material

Supplementary material is available at American Journal of Respiratory Cell and Molecular Biology online.

Conflicts of interest

Please see the ICMJE disclosure forms, which have been provided as supplementary material. F.F.S. and E.F. are employees of Chiesi Farmaceutici S.p.A., which supported the research work. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be perceived as a potential conflict of interest.

Funding

The study was fully supported by Chiesi Farmaceutici S.p.A. and by the National Recovery and Resilience Plan Complementary Investments (PNC, established with the decree law of May 6, 2021, no. 59, converted by law no. 101 of 2021) in the call for the funding of research initiatives for technologies and innovative trajectories in the health and care sectors (Directorial Decree no. 931 of 06-06-2022)—project no. PNC0000003, AdvaNced Technologies for Human-centrEd Medicine (project acronym: ANTHEM).

Data availability

All data presented in this article can be obtained from the authors for noncommercial use, subject to reasonable request and the completion of required data access agreements. The original code used for analyzing the μCT imaging data is available from the lead contact upon request. Additional details necessary to reanalyze the data reported in this study can also be requested from the corresponding author.