The COVID-19 pandemic has precipitated an unprecedented global health crisis, impacting millions. Extensive research has revealed that a substantial proportion of individuals continue to experience persistent health issues following recovery from acute COVID-19 infection, often with manifestations across multiple organ systems [1–4]. Numerous terms (like long COVID, chronic COVID, long-haul COVID, post-acute sequelae of COVID-19 (PASC), post-COVID syndrome, etc.) have been used to describe this condition [2–6], with Long COVID and PASC emerging as the preferred terms in the recent scientific literature [2–4, 6].

Lungs are the most commonly affected organ by COVID-19 and a significant number of patients continue to have prolonged respiratory issues that warrant close observation with appropriate investigations. A recent meta-analysis has shown that lung parenchymal abnormalities are seen in the pooled frequency of 43.5% on one-year imaging follow-ups [7]. CT chest remains an indispensable follow-up imaging tool with many patients often requiring multiple repeat scans to monitor the disease evolution [2, 5, 7–9]. Since CT involves the use of harmful ionizing radiation that carries the risk of producing double-stranded DNA breaks and heritable genetic mutations; such repeated CT acquisitions can lead to increased cumulative dose with resultant increased risk of radiation-induced health hazards in the exposed individuals [10–13]. The extended imaging follow-ups for the long COVID patients necessitate balancing the benefits of diagnosis with the potential long-term risks of ionizing radiation exposure.

The routinely performed standard-dose CT (SDCT) chest delivers an effective radiation dose between 4-7 millisievert (mSv) to the patient [14, 15]. However, with the recent advances in CT hardware and software, radiation exposure to patients can be significantly contained by using newer CT techniques like low-dose CT (LDCT) and ultra-low-dose CT (ULDCT). The effective dose received by a patient undergoing CT chest with LDCT and ULDCT protocols is in the range of 1–4 mSv, and <1 mSv, respectively; and it can be a futuristic and pragmatic way to minimize the health effects of ionizing radiation [16, 17]. The only apparent limitation of these low dose CT techniques is of compromised image quality, that can affect the diagnostic accuracy; and more research is needed in this direction.

There have been few studies on the utility of LDCT and ULDCT in acute COVID-19 pneumonia that have shown promising results [18–24]. However, to the best of our knowledge, only one study to date has been published in the English literature [25] evaluating the utility of ULDCT in patients with Long COVID/PASC. We share our experience in evaluating and comparing ULDCT chest and SDCT chest in terms of image quality and diagnostic accuracy for detecting the lung changes long COVID. With radiation exposure, comparable to only a couple of chest X-rays, ULDCT chest may prove to be a much safer follow-up imaging tool in patients with long COVID.

CT chest is an invaluable imaging tool to evaluate the sequelae of COVID-19 pneumonia and its complications [2, 5, 7, 29, 30]. However, with the enormous burden of COVID-19 recovered patients presenting with a prolonged course of respiratory illness, and many of them requiring frequent repeat scanning, the increased risk of radiation-induced health hazards always remains in these patients [10–13]. Thus, it became imperative to explore newer CT techniques like LDCT and ULDCT aimed at curtailing the radiation dose while still being able to achieve diagnostic quality images in keeping with the principle of “as low as reasonably achievable” (ALARA) [18–24]. Some authors have even attempted to study the utility of radiation-free imaging modalities like lung MRI, but it is still in the nascent stage with only anecdotal studies done in this direction [31, 32].

Our current study is a prospective study to evaluate the role of ULDCT chest in long COVID that has shown comparable image quality, sensitivity, specificity, and diagnostic accuracy of ULDCT to SDCT chest in non-obese subjects. The mean DLP for SDCT in our patients was 392.7 ± 140 mGy.cm, while for ULDCT, it was 18.2 ± 1.7 mGy.cm and we could achieve a net radiation dose reduction of 94.8% ± 1.7%. We acquired ULDCT at 80 kVp and 25 mA which is similar to the acquisition protocol used by Zarei F et al. [18] in another study evaluating the role of ULDCT in acute COVID-19 pneumonia.

The mean SNR of ULDCT recorded in our study was low (∼36% to that of SDCT), but the overall subjective image quality was comparable to that of SDCT in 80% of the ULDCT scans. These findings are in agreement with a study done by Samir et al. [21] who also reported a low SNR with ULDCT, but an acceptable quality for image interpretation.

The common imaging findings recorded in our study were GGOs (77%), interlobular septal thickening/reticulations (67%), atelectatic/parenchymal bands (63%), nodules (26%), and lymphadenopathy (23%). These findings align with the previous studies done on long COVID patients by various authors [2, 5, 7, 9, 33]. However, few patients in our cohort also showed hydropneumothorax, cavitation and tree-in-bud pattern, which can be attributed to the secondary bacterial or fungal infections [29, 30, 34]. The CT severity score calculated on SDCT and ULDCT in our cohort showed a very strong Pearson’s correlation coefficient of 0.996 (p-value < 0.0001), which is concurrent with the study done by Bahrami-Motlagh et al [20].

ULDCT in our study could not reliably depict GGOs in 21 subjects (completely missed in 12 patients and misinterpreted as consolidation in 9 patients). The sensitivity and diagnostic accuracy of ULDCT for detecting GGOs in our study were 72.7% and 79% respectively, which is in concordance with the study done by Zarei et al. [18] who reported a sensitivity of 62%, and diagnostic accuracy of 77%. However, ULDCT achieved a specificity of 100% in our study for detecting GGOs, which is significantly different from the study by Zarei et al. [18], who reported a specificity of 66%.

ULDCT also missed a few other abnormalities in some patients viz., septal thickening/reticulation (n = 19), atelectatic bands (n = 12), bronchiectasis (n = 5), and nodules (n = 6). However, ULDCT in our study achieved an overall diagnostic accuracy of 95% (90.7%–99.4%), which is in concordance with a previous comparative study done by Samir et al. [21] in acute COVID-19 pneumonia who reported similar diagnostic accuracy of ULDCT (90.38%–93.84%). Also, the overall specificity of ULDCT in detecting imaging correlates of long COVID was found to be high (99.4%) in our patient population. These findings suggest that ULDCT has the potential to become a valuable tool in monitoring the radiological progression in long COVID patients.

Another unique feature of our study was that we recorded the BMI of all participants and studied its effect on the image quality of ULDCT. We found that ULDCT scans in obese category patients (BMI ≥30 kg/m²) showed evidence of increased image graininess, increased artifacts, and reduced sharpness which compromised the overall diagnostic performance of ULDCT. Interestingly, this deterioration increased with increasing BMI and was more conspicuous in class II and class III categories of obese patients. In our study cohort, 7% of ULDCT scans showed high levels of image graininess, while blurring, and images of poor diagnostic confidence which were likely to misinterpret/miss imaging findings were seen in 9% each. Interestingly, only two ULDCT scans with compromised image quality belonged to non-obese patients, while all remaining scans in this category belonged to obese subjects. Sharp reproduction of segmental bronchi and lung parenchyma was not seen in 16 and 11 patients respectively, in the obese category. Artifacts were observed in 12 of our obese patients, while ULDCT scans showed high levels of image graininess and poor diagnostic confidence in 5 and 7 obese patients respectively. Likewise, of the 21 ULDCT scans in which GGOs were not detected, 13 of them belonged to the obese category. Furthermore, a statistically significant difference (p-value <0.0001) was observed in the diagnostic performance of ULDCT between obese and non-obese patients. In the obese group, sensitivity and diagnostic accuracy were demonstrably lower, reaching 60.3% (95% CI: 51.2%–68.9%) and 84.9% (95% CI: 77.8%–92.1%), respectively. Conversely, non-obese patients exhibited considerably higher sensitivity (88%, 95% CI: 83.4%–91.7%) and diagnostic accuracy (97.7%, 95% CI: 94.7%–100%). These results of ULDCT hold promise for the future clinical adoption of ULDCT, at least in the non-obese population group. Moreover, these results imply that the ULDCT scan protocol will require further modification (tweaking of the mAs or kVp) in obese patients to achieve desirable results. More research with larger patient cohorts is however required to further understand the effect of BMI on ULDCT scanning.

In the only study published on the utility of ULDCT in Long COVID by Wassipaul et al [25], the authors employed 100 kVp and 50 mAs as scan parameters, while we used 80 kVp and 25 mAs in our study. Despite utilizing comparatively higher scan parameters, Wassipaul’s study reported a lower median DLP (12.6 mGycm) and effective radiation dose (0.1764 mSv) compared to 18.2 mGycm and 0.25 mSv, respectively in our cohort. This can be due to the inherent dose reduction capabilities in some newer CT scanners, as we employed a Philips Brilliance iCT256, while Wassipaul’s utilized a dual-source Siemens Somatom Drive equipped with CAREdose. These findings emphasize the critical role of advanced CT systems with robust dose reduction technologies in minimizing radiation exposure without compromising image quality. However, despite a larger sample size in their study compared to ours (n = 153 vs. n = 100), all (100%) of our patients exhibited at least one relevant radiological finding, in contrast to only 29.4% in their study. This can be attributed to two factors - one, our study cohort included only the patients who were hospitalised with moderate to severe illness during their acute phase, and two the relatively older age group in our study population (52.9 ± 12.9 vs. 47.4 ± 15.3 years), both the recognised risk factors for long COVID [2, 35, 36]. However, our study demonstrated lower sensitivity (78.9% vs. 87.2%), potentially due to a higher obesity rate (21%) and use of lower kVp/mAs settings in our cohort, but it exhibited superior specificity (99.4% vs. 94.9%) and diagnostic accuracy (95% vs. 92.6%). Furthermore, the absence of a comprehensive qualitative (image noise, sharpness, artifacts, European image quality guidelines) and quantitative (including SNR) assessment, and non-inclusion of patient BMI, limits the impact of study by Wassipaul et al.

Despite the American Thoracic Society recommending against the use of ULDCT in ILD [37], we made an initial endeavour to evaluate its utility for assessing lung abnormalities in long COVID. Few other authors have also studied the utility of LDCT and ULDCT chest in the follow-up of patients with ILD with variable results. In a study by Hata et al. [38], it was found that the image quality of ULDCT with model-based iterative reconstruction (MBIR) was worse compared to SDCT, and detailed evaluation of ILD would not be possible with ULDCT. While, another study by Lim et al. [39] showed a comparable diagnostic performance of LDCT (with MBIR) to SDCT in the evaluation of ILDs, but they didn’t evaluate the utility of ULDCT. Since patients with long COVID can show findings similar to those seen in ILD like GGOs, septal thickening, parenchymal bands, etc. [2, 5, 7, 33]; more studies in long COVID patients using ULDCT are needed to substantiate our results.

There were a few limitations in our study. First, the single-center design with a relatively small sample size may have compromised the study’s statistical power. Second, even though we studied the imaging findings in long COVID patients comparing SDCT with ULDCT, we did not use any predefined follow-up timeline to scan the patients and this might have led to a varied spectrum of imaging findings. Third, the inclusion of patients with moderate to severe acute illness might have precluded the identification of more subtle lung abnormalities, potentially introducing bias. Likewise, we used same and fixed CT acquisition protocols for all the patients (both obese and non-obese categories) and that could have also affected the results in our study. Furthermore, there was lack of histopathological correlation to definitively confirm the nature of the observed imaging findings.

In conclusion, the ULDCT chest has the potential to deliver acceptable diagnostic quality images in non-obese long COVID subjects at a much-reduced radiation exposure. The feasibility of ULDCT lies in the fact that it can be done on any CT machine by just changing the kVp and mAs without the need for any additional CT software or hardware. The accuracy of ULDCT can be further improved by taking into account the BMI of the patients and tweaking the scan acquisition protocols accordingly. However, to standardize and optimize the ULDCT protocol and validate its diagnostic accuracy, further multi-institutional research is needed amongst larger cohorts of patients.

This work represents an advance in biomedical science because ULDCT demonstrated comparable diagnostic performance to SDCT in identifying lung abnormalities in long COVID, in non-obese individuals, with significantly reduced radiation dose.