Interstitial lung diseases (ILDs) are a group of over 200 different diseases of the lungs that involve different extents of inflammation, injury to the alveoli, and gradual scarring of the lung tissue.[1] ILDs are a significant contributor to chronic respiratory morbidity and mortality, and are becoming more recognized because of advances in imaging and diagnostics.[2] Early detection is important as some ILDs can progress to irreversible pulmonary fibrosis, respiratory failure, and significant reduction in quality of life.[3] Current strategies focus on identification of interstitial abnormalities prior to progression to more advanced structural and functional changes.[4, 5]

The central role of diagnosis and characterization of ILD is played by high-resolution computed tomography (HRCT) of the chest, which can reveal subtle reticular abnormalities, ground-glass opacities, interlobular septal thickening, traction bronchiectasis, honeycombing, and other characteristic patterns that may not be seen on conventional imaging.[6] Thinner-section CT is also better at identifying early interstitial abnormalities than thicker-section CT. As a result, HRCT is the main imaging reference standard and is the basis for the evaluation of new non-invasive techniques.[7] Repeated HRCT examinations have, however, the disadvantages of exposure to ionizing radiation, and availability, costs, and access to CT may restrict its use, especially in resource-limited healthcare environments.[8]

Lung ultrasound (LUS) is a potentially useful bedside imaging tool for determining interstitial abnormalities.[9] LUS doesn't require any radiation, is portable, relatively inexpensive, repeatable, and can be carried out at the bedside of the patient.[10] Increased B-lines and abnormalities of the pleural line are the main sonographic findings associated with ILD. B-lines may be representative of interstitial involvement, and an irregular, thickened, fragmented, or nodular pleural line may offer further diagnostic clues.[11] A number of clues from the past few years have indicated that LUS is a very sensitive diagnostic tool for ILD, though the specificity depends on the ultrasound protocol, the ultrasound threshold, patient population, and operator experience.[12, 13]

Current available diagnostic evidence is promising but is quite heterogeneous. A systematic review and meta-analysis of nine studies and 888 patients found a pooled sensitivity of 93% and specificity of 61% for B-line-based LUS assessment of ILD and a summary receiver operating characteristic area of 0.91.[14]  In a more recent summary of evidence provided in the ERS/EULAR clinical practice guideline, the authors report a sensitivity of 94% and a specificity of 86% for LUS in 12 observational cohorts of 971 patients with connective-tissue-disease-associated ILD compared with HRCT. However, the guideline advised against using LUS as an alternative to HRCT, citing the criteria of study quality, inconsistent scanning protocols, false-positive and false-negative results, and lack of generalizability of existing study findings.[15] A clinical diagnostic study has also shown high sensitivity of B-line-based criteria, with the total B-line score having 97.4% sensitivity and 97.9% negative predictive value for the presence of ILD, and pleural abnormalities being significantly associated with ILD.[16]

While these encouraging results are present, an important diagnostic question that remains is whether LUS could be used routinely to diagnose early ILD in a wider clinical population, one that is not heavily skewed towards selected connective-tissue-disease patient groups. B-lines do not necessarily indicate ILD and can also be seen in pulmonary oedema, pulmonary infection and other diffuse lung diseases; the interpretation of pleural abnormalities is somewhat operator-dependent. There has been a recent international consensus effort to standardize assessment of B-lines and PLEL irregularity, but moderate, not perfect, inter-reader agreement demonstrates a need for additional clinical validation.[12, 13]  Hence, LUS can be directly compared with HRCT, based on diagnostic accuracy measures including sensitivity, specificity, positive and negative predictive values, overall accuracy, and area under a receiver operating characteristic curve, to determine its role in early detection of ILD.

Early ILD diagnosis is useful clinically, but may not be possible in settings where access to HRCT, cost, radiation exposure, or repeated imaging is an issue. A quick, non-radiographic method that is capable of accurately determining which patients need definitive HRCT assessment would enhance the diagnostic pathway, especially in clinics where advanced imaging technology is not available. Therefore, the present study was performed to evaluate the diagnostic value of lung ultrasound against HRCT in the diagnosis of early interstitial lung disease. This study aims to find the sensitivity, specificity, positive predictive value, negative predictive value, diagnostic accuracy, and ROC-AUC of LUS for the detection of early ILD as compared to HRCT as the reference standard.

A cross-sectional diagnostic accuracy study was carried out for 6 months, from 1st December, 2025 to 30th May, 2026, at the Department of Pulmonology, Khalifa Gul Nawaz Teaching Hospital, Medical Teaching Institution (MTI) Bannu, Khyber Pakhtunkhwa, Pakistan. The sample size was determined in OpenEpi using the approach of diagnostic accuracy. The calculation was based on the lung ultrasound's sensitivity to detect ILD, which was assumed to be 91.3% from a recent hospital-based diagnostic accuracy study, in which LUS was compared with HRCT.[17] The sample size was approximately 159 patients based on sensitivity, assuming a 95% confidence level, a 30% expected prevalence of ILD among all those evaluated for suspected disease, and an absolute precision of 8%. The requirement was not specific; hence, the sensitivity-based estimate was used. After allowing approximately 10% for incomplete data, technically inadequate ultrasound examinations, or failure to undergo HRCT, the final sample size was increased to 176 participants. A consecutive non-probability sampling technique was used. Patients of either sex aged 18 years or older who were referred for evaluation of suspected early interstitial lung disease on the basis of respiratory symptoms, clinical examination, abnormal pulmonary function testing, or an initial radiological suspicion were included. Patients were eligible if clinically indicated and available for LUS; HRCT of the chest was clinically indicated and performed during the study period. Written informed consent was obtained, and these patients were included. Patients with a known advanced ILD diagnosis but not those with a definite radiological diagnosis were excluded, as the study aimed for early diagnosis of these patients. Patients who were expected to have significant abnormalities of LUS B-lines or pleural-line abnormalities, such as those with acute pulmonary edema, extensive pneumonia, pleural effusion, or pneumothorax, were excluded. Patients who could not be scanned by HRCT and other patients who could not have a CT examination because of contraindications or refused were also excluded. Patients who had incomplete LUS or HRCT examinations, or technically inadequate images that were not interpretable, were not included in the final diagnostic analysis. Eligible participants were then consecutively recruited after obtaining ethical approval from the institutional review committee and written informed consent. Data collection was documented on a structured proforma, and demographic and clinical data such as age, sex, presenting respiratory symptoms, duration of symptoms, smoking history, occupational and environmental factors, history of connective-tissue disease, and relevant comorbidities were recorded. Other findings from clinical examination and available pulmonary function tests, if conducted as part of routine clinical assessment, were also recorded. Lung ultrasound was done with a portable ultrasound machine fitted with an appropriate low-frequency convex and/or high-frequency linear transducer. Systematic examination was done bilaterally, anterior, lateral, and posterior chest. B-lines were noted and mapped, as well as any pleural-line abnormalities including pleural thickening, irregularity, fragmentation, or subpleural abnormalities. The ultrasound examination was done before the HRCT findings were reviewed, and the sonographer was blinded to the final HRCT interpretation for standardization. The examination was found to be positive or negative for suspected interstitial involvement on the basis of a predefined LUS criterion, which has been formulated mainly based on the presence and distribution of multiple B-lines along with pleural-line abnormalities. Then HRCT of the chest was done using thin-section high-resolution imaging as per the radiological protocol in the hospital. A radiologist with experience in thoracic imaging who was blinded to the LUS findings interpreted the HRCT images. Ground-glass opacity, reticular abnormalities, thickening of the interlobular septa and traction bronchiectasis, architectural distortion, and honeycombing were noted, suggestive of interstitial involvement. The final HRCT was classified as positive or negative for ILD based on specific radiological criteria and was used as the reference standard for ILD against which LUS was compared. Both investigations were performed as close to each other in time as practically possible to minimize changes in the patient's clinical or radiological status. The data gathered were entered into and analyzed using IBM SPSS Statistics (version 26). Continuous variables such as age and other quantitative measurements that were normally distributed were presented as mean ± SD, while those variables that were not normally distributed were presented as median (IQR). Data for categorical variables such as sex, presenting symptoms, smoking status, LUS findings, pleural abnormalities, and HRCT-defined ILD were presented as frequencies and percentages. Normality of continuous variables was tested by the Shapiro–Wilk test. HRCT was used as the reference standard for diagnostic performance assessment of LUS. A 2×2 contingency table was created to categorize the true positive, true negative, false positive, and false negative findings. Sensitivity, specificity, positive predictive value, negative predictive value, overall diagnostic accuracy, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were calculated, with 95% confidence intervals. The receiver operating characteristic (ROC) curve analysis was used to assess the discriminatory power of the LUS B-line assessment, and the area under the ROC curve (AUC) was calculated. If the continuous or ordinal LUS was available, the best diagnostic cut-off was identified using the ROC curve and the Youden index. Agreement was tested between LUS and HRCT classification using Cohen's kappa coefficient. The chi-square test was used to determine association between categorical clinical characteristics and HRCT-documented ILD. The value of P<0.05 was considered as statistically significant.

A total of 176 participants were included in the analysis. The mean age was 49.8 ± 13.7 years, with males comprising 57.4% of the study population. Symptoms involving the respiratory system were more prevalent, including dyspnea and dry cough, whereas connective tissue disease and associated occupational/environmental exposure were less common in participants. (Table 1)

Multiple B-lines were the most common sonographic finding among the abnormalities LUS demonstrated. Abnormalities of the pleura were also detected, such as irregularity, thickening, fragmentation, and subpleural changes. A total of 80 participants had suspected ILD and were classified as LUS-positive. (Table 2)

HRCT was used to detect early ILD in 69 participants. Reticular abnormalities and ground-glass opacities were most common, while traction bronchiectasis, architectural distortion, and honeycombing were less common. (Table 3)

LUS correctly identified most of the participants who had ILD, with relatively few false-negative results, as compared to the reference standard i-e, HRCT. The diagnostic parameters calculated from the images had good sensitivity and negative predictive value and moderate agreement between LUS and HRCT classification. (Table 4)

Several clinical features were significantly related to HRCT-diagnosed ILD. The following variables had statistically significant associations with diagnosis of ILD: older age, male sex, current or previous smoking, connective-tissue disease, occupational/environmental exposures, and fine inspiratory crackles. (Table 5)

All assessed LUS measures showed discriminatory power using ROC analysis. The combination of B-line changes with pleural-line changes gave the best assessment of diagnostic discrimination (AUC of 0.928), and the optimum threshold performance was generally comparable to that of the overall LUS classification. (Table 6)

Table 1. Demographic and clinical characteristics of study participants (n=176)

Table 2. Lung ultrasound findings among study participants (n=176)

Table 3. HRCT findings and final diagnosis of interstitial lung disease (n=176)

Table 4. Comparison of LUS with HRCT for detection of early ILD

Table 5. Association of clinical characteristics with HRCT-confirmed ILD

Table 6. ROC analysis of lung ultrasound for detection of early ILD

In this study, LUS showed high diagnostic accuracy for early ILD diagnosis with HRCT as the reference standard. The sensitivity, specificity, diagnostic accuracy, and NPV were 91.3%, 84.1%, 87.0%, and 93.8%, respectively, and the agreement between the two imaging modalities was substantial (κ=0.73). The ROC analysis also showed the B-line score and combined B-line and pleural-line abnormalities to have excellent discriminatory power, with an AUC of 0.914 and 0.928, respectively. The results are consistent with a possible use of LUS as a highly effective, non-ionizing imaging tool for the initial diagnosis of patients with suspected early ILD. Our sensitivity of 91.3% is close to the sensitivity reported by Fairchild et al. (2021), who tested a pleural-line-based LUS protocol in 20 systemic sclerosis patients and found 100% sensitivity and 82% specificity against HRCT. This is significant because their LUS criteria included pleural irregularity, thickening, and granularity, which are also the same pleural abnormalities that were assessed in the present study. The lower sensitivity observed in our study could be due to the fact that we evaluated early ILD compared to systemic-sclerosis-associated disease, and the broader population.[18] Gargani et al. (2022) evaluated different B-line thresholds and scanning approaches in 69 patients with systemic sclerosis who underwent LUS and HRCT on the same day. They highlighted the role of both B-lines and pleural-line abnormalities in achieving the detection of ILD, and that methodological variations in the area being scanned and the B-line threshold significantly affect in terms of diagnostic performance. This helps the use of a standardized examination that includes both B-lines and pleural abnormalities, rather than using a single sonographic feature.[19] In the same manner, Bruni et al. (2022) showed that the total B-line count has an AUC of 0.85 for the diagnosis of extensive ILD on HRCT in systemic sclerosis patients. They had an AUC lower than the B-line score (0.914) and combined LUS (0.928) in our study. The discrepancy could be due to the different outcome measured: Bruni et al. studied discrimination of extensive disease, while our study assessed early ILD within a clinically suspected population.[20] In the ILD-suspected patient population of 66 subjects in a 2022 observational diagnostic study, the sensitivity of LUS was 93% compared with HRCT, and specificity was 73%. The negative likelihood ratio of LUS was 0.10 (51 of 55 HRCT positive). Our sensitivity was therefore very similar, while our specificity was higher at 84.1%. This may partly be due to LUS being less specific in our study because of the exclusion of conditions with B-lines or pleural abnormalities like pulmonary edema, extensive pneumonia, pleural effusion, and pneumothorax.[21] In 2022, Xie et al. were able to evaluate HRCT as a reference standard in a study of 88 patients with CTD, including 65 with CTD-associated ILD. LUS showed a positive correlation with HRCT findings, with an 86.6% sensitivity and 82.6% specificity. B-line burden, B-line frequency, pleural thickness, and irregularity had a positive correlation with HRCT findings. The direction of the findings was the same, with a slight increase in our sensitivity (91.3%) and specificity (84.1%), especially in the diagnostic value of B-lines and pleural-line abnormalities.[22] A diagnostic study in Nepal in 2023 found a higher level of sensitivity for B-line assessment. The overall best sensitivity and NPV were obtained with a total B-line score ≥ 5, while NPV and specificity were highest with the positive chest-area score. Our sensitivity and NPV were somewhat lower, although still high. The difference could be due to varying populations, disease spectrum, ultrasound protocol, and diagnostic threshold. Importantly, the Nepalese study also demonstrated a significant association between pleural-line abnormalities and ILD, further supporting our use of the combination of LUS abnormalities in the evaluation of ILD.[16] In 14 preliminary scans, a 2024 study compared LUS and CT in systemic sclerosis patients, focusing on qualitative observations of the B-line and pleural line. The study used both a ≥10 B-line and Fairchild pleural-line criteria, both of which are now being recognised as having an important role in the diagnosis of LUS, which depends greatly on the diagnostic criteria used for the sonographic assessment. This methodological approach is similar to our combined assessment, which yielded the highest AUC of 0.928; therefore, the combined assessment of both interstitial artifacts and pleural abnormalities may provide additional information.[23] The findings are also consistent with a 2024 multinational study of asymptomatic patients with rheumatoid arthritis-associated ILD. In that study, 199 patients underwent LUS using a 14-area protocol, with HRCT serving as the reference standard; 53 patients had HRCT-confirmed ILD, and the reported NPV of LUS was 93%. Our NPV came in at 93.8%, which was very close. This is clinically important because a high NPV means that a negative LUS result may help exclude patients with a low likelihood of ILD, but when clinical suspicions are still there, HRCT is required.[24, 25] LUS was studied in patients with CTD-ILD in a 2025 study by Watanabe et al. in Japan, and the total B-line count was found to be correlated with the degree of ILD on HRCT (r = 0.66 in all patients and r = 0.78 in patients with systemic sclerosis). However, the concept that underlying pulmonary involvement is associated with increasing interstitial abnormalities is supported by the high AUC achieved for B-line assessment in our study, although we did not quantify the correlation with extent of involvement on HRCT.[26] A diagnostic-accuracy study in 2025 patients with suspected ILD was performed at Combined Military Hospital Gujranwala, Pakistan. HRCT confirmed 87 patients as having ILD, while LUS was positive in 72 cases. The sensitivity, specificity, and overall diagnostic accuracy reported were 81.6%, 87.5%, and 82.4%, respectively. In our study, sensitivity and overall accuracy were slightly higher, while specificity was slightly lower. The broadly similar findings in two hospital populations in Pakistan lend support to the external validity of LUS in the Pakistani clinical context, and the somewhat different prevalence of disease and patient inclusion may account for some variability in predictive value and sensitivity.[27] Recent results from 2026 confirm the importance of quantitative LUS but point out some of its weak points. Du et al. performed a study on 117 patients with CTD-ILD and discovered a positive relationship between the LUS score and the HRCT Warrick score (rₛ=0.304, p=0.001). LUS, however, discriminated severe disease moderately (AUC 0.642), and this was significantly enhanced by adding DLCO. This is in line with our observation of the high diagnostic discrimination of detecting ILD and indicates that detection and severity stratification may be two distinct clinical applications. The extent and severity of the disease can be estimated by HRCT and pulmonary-function parameters, while interstitial involvement can be identified well on LUS.[28] In conclusion, the present results indicate that LUS can serve as a complementary, but not replacement, imaging modality for ILD. The high sensitivity, NPV, and AUC recorded in this study suggest that LUS can detect a significant number of early stages of ILD without using X-rays and in relatively easily available bedside devices. However, false-negative or false-positive tests have been reported, and HRCT is of major value for definitive diagnosis of interstitial patterns, the range of the disease, fibrosis, and structural abnormalities, which cannot be investigated by ultrasound. In contemporary literature, LUS is also found to be useful for screening, triage, and monitoring purposes and not a complete substitute for HRCT. LIMITATIONS There were a number of limitations in the study. First, the study was performed in a single tertiary care hospital, and the results might not be applicable to other healthcare environments and to other patient groups. Second, there has been a possibility that non-probability sampling has introduced selection bias as the data were collected. Third, although there may be some degree of observer variability in the interpretation of the radiology, HRCT was used as the reference standard. Fourth, LUS is operator-dependent, and variations in ultrasound experience can influence the identification and interpretation of the B line and pleural abnormality. Fifth, a diagnosis exclusion strategy for patients with multiple diseases having similar interstitial sonographic features was undertaken, which could limit the diagnostic specificity in clinical practice. Lastly, the cross-sectional design precluded the capability of assessing the capacity of LUS for monitoring ILD progression over time.

The sensitivity, negative predictive value, and discriminatory ability of lung ultrasound for early ILD detection were high, and very much comparable to HRCT. B-line and pleural-line assessment were both highly diagnostic and showed good agreement with HRCT. LUS may hence be a valuable adjunctive, radiation-free method for initial evaluation of patients with suspected early ILD. Nevertheless, HRCT is still useful for confirmation and characterization of interstitial abnormalities.