A model for classification of invasive fungal rhinosinusitis by computed tomography

Abstract

Our purpose was to classify acute invasive fungal rhinosinusitis (AIFR) caused by Mucor versus Aspergillus species by evaluating computed tomography radiological findings. Two blinded readers retrospectively graded radiological abnormalities of the craniofacial region observed on craniofacial CT examinations obtained during initial evaluation of 38 patients with eventually pathology-proven AIFR (13:25, Mucor:Aspergillus). Binomial logistic regression was used to analyze correlation between variables and type of fungi. Score-based models were implemented for analyzing differences in laterality of findings, including the 'unilateral presence' and 'bilateral mean' models. Binary logistic regression was used, with Score as the only predictor and Group (Mucor vs Aspergillus) as the only outcome. Specificity, sensitivity, positive predictive value, negative predictive value and accuracy were determined for the evaluated models. Given the low predictive value of any single evaluated anatomical site, a 'bilateral mean' score-based model including the nasal cavity, maxillary sinuses, ethmoid air cells, sphenoid sinus and frontal sinuses yielded the highest prediction accuracy, with Mucor induced AIFR correlating with higher prevalence of bilateral findings. The odds ratio for the model while integrating the above anatomical sites was 12.3 (p < 0.001). PPV, NPV, sensitivity, specificity and accuracy were 0.85, 0.82, 0.92, 0.69 and 0.84 respectively. The abnormal radiological findings on craniofacial CT scans of Mucor and Aspergillus induced AIFR could be differentiated based on laterality, with Mucor induced AIFR associated with higher prevalence of bilateral findings.

Figure 1

Examples of findings in patients with Mucor AIFR. (A) Coronal CT image shows bilateral mucosal thickening involving the maxillary sinuses (asterisks) and the nasal cavity. (B) Axial image in the same patient demonstrates bilateral mucosal thickening of ethmoid air cells (asterisk). (C) Coronal image in a different patient shows bilateral maxillary sinuses involvement and left orbital involvement with fatty infiltration of the left medial and inferior extraconal orbital fat (arrows). (D) Coronal CT of a different patient shows bilateral mucosal thickening of the maxillary sinuses with bony dehiscence along the inferior aspect of the right maxillary sinus.

Figure 2

Examples of findings in patients with Aspergillus AIFR. (A) Coronal CT image shows unilateral mucosal thickening of the right maxillary sinus (asterisk) and the ethmoid air cells. (B, C) Axial images from the same patient shows fatty infiltration of the anterior periantral fat and the posterior periantral fat, without bony dehiscence. (D) Another patient with Aspergillus AIFR with unilateral mucosal thickening of the maxillary sinus and the nasal cavity (asterisk).

Figure 3

Conditional density diagram of the bilateral mean model. Patients with AIFR caused by Mucor species had higher mean score values compared with patients with AIFR caused by Aspergillus species.