Our results indicate that while baseline CT and laboratory parameters are insufficient to predict the PET-based metabolic activity of COVID-19 pneumonia measured a median of seven days after hospital admission, the percentage of all CT lung abnormalities, IL-6, and LDH levels at follow-up examination provide helpful insights regarding the lungs’ inflammatory status. To the best of our knowledge, the current investigation is the first to assess the total inflammatory activity value in COVID-19-pneumonia. In the present study, TIA did not show markedly different correlation patterns to CT and laboratory parameters than MIV, however, the combined evaluation of metabolic extent and activity of COVID-19-pneumonia by TIA values may add further information to the understanding the occurrence of post-COVID-19 residual lung lesions.
Similarly to other microbial lung infections, acute-phase COVID-19-pneumonia presents with increased uptake in lung areas with active inflammation [10, 14, 15, 20]. Growing number of publications show a heterogeneous FDG-uptake and varied extent of active pneumonia among COVID-19 patients. Lung SUVmax was investigated by Dietz et al in their study of 13 patients with COVID-19-pneumonia, the value ranging from 4.7 to 16.3 which is comparable to the range in our study of 2.1–12.2 [14]. Furthermore, as a more reliable indicator of inflammation, we also investigated the MIV and the TIA, the latter as the product of MIV and mean SUV in the lung VOI, which both served as the endpoints when evaluating CT and laboratory parameters. MIV was included in the study of Dietz et al as “lung hypermetabolic volume” (using SUV cut-off of 4 for segmentation), ranging from 3.6 ml to 990 ml, while in our study the range of MIV (using SUV cut-off of 2 for segmentation) was between 1 ml and 1797 ml [14]. We also investigated the reproducibility of MIV segmentations, and an excellent correlation was shown between the calculations of two independent readers.
CT severity score is an established tool in reporting CT scans of COVID-19 patients [18, 21]. In our study, its value at baseline CT could not accurately predict the metabolic extent of active inflammation by the PET examination (MIV), performed a median of seven days later. However, the CT severity scores on the follow-up CT showed better correlation with MIV values of the PET component.
Software tools based on DL can be utilized to automatically or semi-automatically segment lung lesions in COVID-19 pneumonia [17, 22–24]. We implemented a DL-based algorithm for quantitative assessment of pneumonia, developed at the Cedars-Sinai Medical Center (Los Angeles, CA, USA) [17]. Correlation of DL-calculated CT inflammatory volumes and metabolic PET-parameters (MIV, TIA) were low in case of baseline CT and high in case of follow-up CT (as part of the PET/CT scan), but even in the latter case, marked differences were observed in the absolute volumetric values measured on the CT and the PET component.
Several publications indicate the relationship between elevated levels of CRP, LDH, and IL-6 and the clinical severity of COVID-19-pneumonia, with the first two parameters showing correlation with the extent of lung lesions on CT [25–28]. In our patient cohort, only the LDH levels at the time of the PET/CT yielded good correlation with the metabolic activity and extent of COVID-19-pneumonia, whereas baseline and follow-up IL-6 levels could be utilized in regression models.
This combination of laboratory and CT parameters – which are more widely available than PET/CT in acute infectious status – helps to better determine the inflammatory activity of COVID-19-pneumonia, as certain discrepancies frequently occur between the extent of metabolic and morphological lung lesions (Fig. 3–4).
Our study has some limitations. Despite following a pre-set patient inclusion algorithm, some spread in the intervals between baseline and follow-up scanning can be observed, coupled with the difficulty of defining the starting point of the patients’ coronavirus disease. These can result in the PET-imaging of patients in slightly different points in the course of COVID-19-pneumonia (pre-peak, peak, post-peak). Apart from individual variabilities and the inherent differences between morphological and metabolic imaging, the lack of synchronising of PET-studies at the peak disease activity may explain the moderate correlation between CT and PET volumes.
Due to ethical and scheduling considerations, patients in need of mechanical ventilation support were not selected for the study, thus excluding a patient group where medical research in early prediction of severe disease would be highly relevant. Furthermore, as all patients in this study received a specific, standard treatment, our results may not be applicable to patients with different therapeutic management - however, this homogenous patient selection strengthens the reliability of the relationship between morphological, laboratory and metabolic findings. The single-centre setup and relatively small cohort of patients may also limit the applicability of our results in different settings.
Our planned further research in this cohort is to evaluate the effect of PET-metabolic activity on residual scarring in the lung and on patient life quality.