High CT attenuation values relative to the brainstem may predict squamous cell carcinoma arising from inverted papilloma

Deep learning model for differentiating nasal cavity masses based on nasal endoscopy images

BACKGROUND

Nasal polyps and inverted papillomas often look similar. Clinically, it is difficult to distinguish the masses by endoscopic examination. Therefore, in this study, we aimed to develop a deep learning algorithm for computer-aided diagnosis of nasal endoscopic images, which may provide a more accurate clinical diagnosis before pathologic confirmation of the nasal masses.

METHODS

By performing deep learning of nasal endoscope images, we evaluated our computer-aided diagnosis system's assessment ability for nasal polyps and inverted papilloma and the feasibility of their clinical application. We used curriculum learning pre-trained with patches of nasal endoscopic images and full-sized images. The proposed model's performance for classifying nasal polyps, inverted papilloma, and normal tissue was analyzed using five-fold cross-validation.

RESULTS

The normal scores for our best-performing network were 0.9520 for recall, 0.7900 for precision, 0.8648 for F1-score, 0.97 for the area under the curve, and 0.8273 for accuracy. For nasal polyps, the best performance was 0.8162, 0.8496, 0.8409, 0.89, and 0.8273, respectively, for recall, precision, F1-score, area under the curve, and accuracy. Finally, for inverted papilloma, the best performance was obtained for recall, precision, F1-score, area under the curve, and accuracy values of 0.5172, 0.8125, 0.6122, 0.83, and 0.8273, respectively.

CONCLUSION

Although there were some misclassifications, the results of gradient-weighted class activation mapping were generally consistent with the areas under the curve determined by otolaryngologists. These results suggest that the convolutional neural network is highly reliable in resolving lesion locations in nasal endoscopic images.

High CT Attenuation Values Relative to the Brainstem Predict Fungal Hyphae Within the Sinus

Objectives

There is currently no established objective diagnostic indicator for the differentiation of sinus fungal ball (SFB) from unilateral nonfungal chronic sinusitis (UCRS). This study evaluated whether computed tomography (CT) attenuation values relative to those of the brainstem (relative CT number) are useful for differentiating SFB from UCRS.

Materials and Methods

Consecutive patients who were pathologically diagnosed with SFB or UCRS between 2013 and 2021 were retrospectively identified. The relative CT numbers of region of interest (ROIs) within the sinuses were compared between the two patient groups. Factors with predictive power for differentiating SFBs from UCRSs were identified by uni/multivariable logistic regression analyses.

Results

One hundred and eighty-three patients with unilateral chronic sinusitis were finally analyzed (SFB, 86 cases; UCRS, 97 cases). Regardless of the presence or absence of calcified lesions, the relative CT numbers in SFB were significantly higher than those in UCRS. ROIs showing high relative CT numbers were those where fungal hyphae were present. In the uni/multivariable logistic regression analysis, age (p < 0.001), relative CT number (p < 0.001), and calcification (p = 0.002) had predictive value for distinguishing SFB from UCRS. Within those cases not showing calcification, age (p = 0.004) and relative CT number (p < 0.001) were predictive factors for differentiating SFB from UCRS. A relative CT number >1.5 was significantly associated with SFB (sensitivity, 70%; specificity, 91%), with a significantly larger area under the receiver operating characteristics curve than age.

Conclusions

High relative CT numbers within the sinus are strongly associated with the presence of fungal hyphae, and measurement of relative CT number is a powerful adjunctive diagnostic method for distinguishing between SFB and UCRS.

High CT values relative to the brainstem differentiate inverted papillomas from nasal polyps

OBJECTIVES

A diagnostic indicator for the differentiation of inverted papillomas (IPs) from inflammatory nasal polyps (NPs) has not been established. This study aimed to evaluate whether CT attenuation values relative to those of the brainstem (relative CT number) could be useful for differentiating IPs from NPs.

MATERIAL AND METHODS

Consecutive patients who were pathologically diagnosed with IP or NP between 2005 and 2019 were retrospectively identified. Relative CT numbers were compared between the two patient groups. The factors with predictive power for differentiating IPs from NPs were identified by univariate and multivariate logistic regression analyses.

RESULTS

One hundred and twenty-two sinonasal masses were finally analysed (IP, 51 cases; NP, 71 cases). Relative CT numbers were significantly higher in IP than in NP (P < 0.001). Univariate logistic regression analysis showed relative CT number, bone erosion and bone thickening to have predictive value for differentiating IPs from NPs (relative CT number, P < 0.001; bone erosion, p = 0.04; bone thickening, P < 0.001). In the multivariate logistic regression analysis, relative CT number and bone thickening had predictive value for distinguishing IP from NP (relative CT number, p < 0.001; bone thickening, p = 0.02). The optimum cut off value calculation from the area under the receiver operating characteristics curve indicated that a relative CT number >1.3 was significantly associated with IP (sensitivity, 72.6%; specificity, 87.3%). Within cases not showing bone thickening, only the relative CT number was a predictive factor for differentiating IPs from NPs in the univariate analysis.

CONCLUSIONS

High relative CT numbers could potentially indicate IP, and their measurement could provide a basis for differentiating IPs from NPs.