Artificial intelligence for cellular phenotyping diagnosis of nasal polyps by whole-slide imaging

Multiple instance learning for eosinophil quantification of sinonasal histopathology images: A hierarchical determination on whole slide images

KEY POINTS

We proposed a hierarchical framework including an unsupervised candidate image selection and a weakly supervised patch image detection based on multiple instance learning (MIL) to effectively estimate eosinophil quantities in tissue samples from whole slide images. MIL is an innovative approach that can help deal with the variability in cell distribution detection and enable automated eosinophil quantification from sinonasal histopathological images with a high degree of accuracy. The study lays the foundation for further research and development in the field of automated histopathological image analysis, and validation on more extensive and diverse datasets will contribute to real-world application.

[Artificial intelligence-assisted prediction of olfactory disorders in patients with chronic rhinosinusitis]

Objective:To analyze the influencing factors and perform the prediction of olfactory disorders in patients with chronic rhinosinusitis（CRS） based on artificial intelligence. Methods:The data of 75 patients with CRS who underwent nasal endoscopic surgery from October 2021 to February 2023 in the Department of Otorhinolaryngology Head and Neck Surgery, the Third Affiliated Hospital of Sun Yat-sen University were analyzed retrospectively. There were 53 males and 22 females enrolled in the study, with a median age of 42.0 years old. The CRS intelligent microscope interpretation system was used to calculate the proportion of area glands and blood vessels occupy in the pathological sections of each patient, and the absolute value and proportion of eosinophils, lymphocytes, plasma cells and neutrophils. The patients were grouped according to the results of the Sniffin' Sticks smell test, and the clinical baseline data, differences in nasal mucosal histopathological characteristics, laboratory test indicators and sinus CT were compared between the groups. Determine the independent influencing factors of olfactory disorders and receiver operating characteristic curves（ROC） were used to evaluate the performance of the prediction model. Statistical analysis was performed using SPSS 25.0 software. Results:Among the 75 CRS patients, 25 cases（33.3%） had normal olfaction and 50 cases（66.7%） had olfactory disorders. Multivariate Logistic regression analysis showed that tissue eosinophils percentage（OR=1.032, 95%CI 1.002-1.064, P=0.036）, Questionnaire of olfactory disorders-Negative statement（QOD-NS）（OR=1.079, 95%CI 1.004-1.160, P=0.040） and Anterior olfactory cleft score（AOCS）（OR=2.672, 95%CI 1.480-4.827, P=0.001） were independent risk factors for olfactory disorders in CRS patients. Further research found that the area under the ROC curve（AUC） of the combined prediction model established by the tissue eosinophil percentage, QOD-NS and AOCS was 0.836（95%CI 0.748-0.924, P<0.001）, which is better than the above single factor prediction model in predicting olfactory disorders in CRS. Conclusion:Based on pathological artificial intelligence, tissue eosinophil percentage, QOD-NS and AOCS are independent risk factors for olfactory disorders in CRS patients, and the combination of the three factors has a good predictive effect on CRS olfactory disorders.

Advances in Image-Based Artificial Intelligence in Otorhinolaryngology-Head and Neck Surgery: A Systematic Review

OBJECTIVE

To update the literature and provide a systematic review of image-based artificial intelligence (AI) applications in otolaryngology, highlight its advances, and propose future challenges.

DATA SOURCES

Web of Science, Embase, PubMed, and Cochrane Library.

REVIEW METHODS

Studies written in English, published between January 2020 and December 2022. Two independent authors screened the search results, extracted data, and assessed studies.

RESULTS

Overall, 686 studies were identified. After screening titles and abstracts, 325 full-text studies were assessed for eligibility, and 78 studies were included in this systematic review. The studies originated from 16 countries. Among these countries, the top 3 were China (n = 29), Korea (n = 8), the United States, and Japan (n = 7 each). The most common area was otology (n = 35), followed by rhinology (n = 20), pharyngology (n = 18), and head and neck surgery (n = 5). Most applications of AI in otology, rhinology, pharyngology, and head and neck surgery mainly included chronic otitis media (n = 9), nasal polyps (n = 4), laryngeal cancer (n = 12), and head and neck squamous cell carcinoma (n = 3), respectively. The overall performance of AI in accuracy, the area under the curve, sensitivity, and specificity were 88.39 ± 9.78%, 91.91 ± 6.70%, 86.93 ± 11.59%, and 88.62 ± 14.03%, respectively.

CONCLUSION

This state-of-the-art review aimed to highlight the increasing applications of image-based AI in otorhinolaryngology head and neck surgery. The following steps will entail multicentre collaboration to ensure data reliability, ongoing optimization of AI algorithms, and integration into real-world clinical practice. Future studies should consider 3-dimensional (3D)-based AI, such as 3D surgical AI.

Machine learning method for the cellular phenotyping of nasal polyps from multicentre tissue scans

BACKGROUND

This study aimed to establish a convenient and accurate chronic rhinosinusitis evaluation platform CRSAI 1.0 according to four phenotypes of nasal polyps.

RESEARCH DESIGN AND METHODS

Tissue sections of a training (n = 54) and test cohort (n = 13) were sourced from the Tongren Hospital, and those for a validation cohort (n = 55) from external hospitals. Redundant tissues were automatically removed by the semantic segmentation algorithm of Unet++ with Efficientnet-B4 as backbone. After independent analysis by two pathologists, four types of inflammatory cells were detected and used to train the CRSAI 1.0. Dataset from Tongren Hospital were used for training and testing, and validation tests used the multicentre dataset.

RESULTS

The mean average precision (mAP) in the training and test cohorts for tissue eosinophil%, neutrophil%, lymphocyte%, and plasma cell% was 0.924, 0.743, 0.854, 0.911 and 0.94, 0.74, 0.839, and 0.881, respectively. The mAP in the validation dataset was consistent with that of the test cohort. The four phenotypes of nasal polyps varied significantly according to the occurrence of asthma or recurrence.

CONCLUSIONS

CRSAI 1.0 can accurately identify various types of inflammatory cells in CRSwNP from multicentre data, which could enable rapid diagnosis and personalized treatment.

A novel inflammatory endotype diagnostic model based on cytokines in chronic rhinosinusitis with nasal polyps

Background

Type 2 CRSwNP is characterized by severe symptoms, multiple comorbidities, longer recovery course and high recurrence rate. A simple and cost-effective diagnostic model for CRSwNP endotype integrating clinical characteristics and histopathological features is urgently needed.

Objective

To establish a clinical diagnostic model of inflammatory endotype in CRSwNP based on the clinical characteristics, pathological characteristics, and cytokines profile in the polyp tissue of patients.

Methods

A total of 244 participants with CRSwNP were enrolled at 2 different centers in China and Belgium from 2018 to 2020. IL-5 level of nasal polyp tissue was used as gold standard. Clinical characteristics were used to establish diagnostic models. The area under the receiver operating curve (AUC) was used to evaluate the diagnostic performance. The study was approved by the ethics board of the First Affiliated Hospital of Sun Yat-sen University ([2020] 302), and written informed consent was obtained from all subjects before inclusion.

Results

In total, 134 patients from China (training set) and 110 patients from Belgium (validation set) were included. The logistic regression (LR) model in predicting inflammatory endotype of CRSwNP showed the AUC of 83%, which was better than the diagnostic performance of machine learning models (AUC of 61.14%-82.42%), and single clinical variables. We developed a simplified scoring system based on LR model which shows similar diagnostic performance to the LR model (P = 0.6633).

Conclusion

The LR model in this diagnostic study provided greater accuracy in prediction of inflammatory endotype of CRSwNP than those obtained from the machine learning model and single clinical variable. This indicates great potential for the use of diagnostic model to facilitate inflammatory endotype evaluation when tissue cytokines are unable to be measured.

A review of current biomarkers in chronic rhinosinusitis with or without nasal polyps

INTRODUCTION

Chronic rhinosinusitis (CRS) is a heterogeneous disease with a variety of cellular and molecular pathophysiologic mechanisms. Biomarkers have been explored in CRS using various phenotypes, such as polyp recurrence after surgery. Recently, the presence of regiotype in CRS with nasal polyps (CRSwNP) and the introduction of biologics for the treatment of CRSwNP has indicated the importance of endotypes, and there is a need to elucidate endotype-based biomarkers.

AREAS COVERED

Biomarkers for eosinophilic CRS, nasal polyps, disease severity, and polyp recurrence have been identified. Additionally, endotypes are being identified for CRSwNP and CRS without nasal polyps using cluster analysis, an unsupervised learning technique.

EXPERT OPINION

Endotypes in CRS have still being established, and biomarkers capable of identifying endotypes of CRS are not yet clear. When identifying endotype-based biomarkers, it is necessary to first identify endotypes clarified by cluster analysis for outcomes. With the application of machine learning, the idea of predicting outcomes using a combination of multiple integrated biomarkers, rather than a single biomarker, will become mainstream.

Endotyping Difficult-to-Treat Chronic Rhinosinusitis with Nasal Polyps by Structured Histopathology

INTRODUCTION

This study aimed to identify the histopathologic characteristics associated with difficult-to-treat chronic rhinosinusitis with nasal polyps (CRSwNPs), enabling physicians to predict the risk of poor outcome after endoscopic sinus surgery (ESS).

METHODS

A prospective cohort study performed at the First Affiliated Hospital of Sun Yat-sen University between January 2015 and December 2018 with CRSwNP patients who underwent ESS. Polyp specimens were collected during surgery and were subjected to structured histopathological evaluation. Difficult-to-treat CRSwNPs were determined at 12-15 months post-operation according to the European Position Paper. Multiple logistic regression model was used to assess the association between histopathological parameters and the difficult-to-treat CRSwNP.

RESULTS

Among 174 subjects included in the analysis, 49 (28.2%) were classified with difficult-to-treat CRSwNP, which had higher numbers of total inflammatory cells, tissue eosinophils, and percentages of eosinophil aggregates and Charcot-Leyden crystals (CLC) formation but a lower number of interstitial glands than the nondifficult-to-treat CRSwNP. Inflammatory cell infiltration (adjusted OR: 1.017), tissue eosinophilia (adjusted OR: 1.005), eosinophil aggregation (adjusted OR: 3.536), and CLC formation (adjusted OR: 6.972) were independently associated with the difficult-to-treat outcome. Furthermore, patients with tissue eosinophil aggregation and CLC formation had an increasingly higher likelihood of uncontrolled disease versus those with tissue eosinophilia.

CONCLUSION

The difficult-to-treat CRSwNP appears to be characterized by increased total inflammatory infiltrates, tissue eosinophilia, eosinophil aggregation, and CLC formation in structured histopathology.

Deep learning radiomics-based preoperative prediction of recurrence in chronic rhinosinusitis

Chronic rhinosinusitis (CRS) is characterized by poor prognosis and propensity for recurrence even after surgery. Identification of those CRS patients with high risk of relapse preoperatively will contribute to personalized treatment recommendations. In this paper, we proposed a multi-task deep learning network for sinus segmentation and CRS recurrence prediction simultaneously to develop and validate a deep learning radiomics-based nomogram for preoperatively predicting recurrence in CRS patients who needed surgical treatment. 265 paranasal sinuses computed tomography (CT) images of CRS from two independent medical centers were analyzed to build and test models. The sinus segmentation model achieved good segmentation results. Furthermore, the nomogram combining a deep learning signature and clinical factors also showed excellent recurrence prediction ability for CRS. Our study not only facilitates a technique for sinus segmentation but also provides a noninvasive method for preoperatively predicting recurrence in patients with CRS.

Artificial intelligence, machine learning, and deep learning in rhinology: a systematic review

PURPOSE

This PRISMA-compliant systematic review aims to analyze the existing applications of artificial intelligence (AI), machine learning, and deep learning for rhinological purposes and compare works in terms of data pool size, AI systems, input and outputs, and model reliability.

METHODS

MEDLINE, Embase, Web of Science, Cochrane Library, and ClinicalTrials.gov databases. Search criteria were designed to include all studies published until December 2021 presenting or employing AI for rhinological applications. We selected all original studies specifying AI models reliability. After duplicate removal, abstract and full-text selection, and quality assessment, we reviewed eligible articles for data pool size, AI tools used, input and outputs, and model reliability.

RESULTS

Among 1378 unique citations, 39 studies were deemed eligible. Most studies (n = 29) were technical papers. Input included compiled data, verbal data, and 2D images, while outputs were in most cases dichotomous or selected among nominal classes. The most frequently employed AI tools were support vector machine for compiled data and convolutional neural network for 2D images. Model reliability was variable, but in most cases was reported to be between 80% and 100%.

CONCLUSIONS

AI has vast potential in rhinology, but an inherent lack of accessible code sources does not allow for sharing results and advancing research without reconstructing models from scratch. While data pools do not necessarily represent a problem for model construction, presently available tools appear limited in allowing employment of raw clinical data, thus demanding immense interpretive work prior to the analytic process.

Neutrophilic inflammation in chronic rhinosinusitis

PURPOSE OF REVIEW

Over the last years, extensive research has been done on neutrophils and their contribution in chronic rhinosinusitis (CRS), and made it clear that they are more than just a bystander in this disease. In this article, we will review all recent publications on this topic and look to what the future hold regarding therapeutics targeting the neutrophilic inflammation in CRS.

RECENT FINDINGS

Evidence is growing that the presence of neutrophils are associated with a worse disease outcome in certain CRS patient groups. They are highly activated in type 2 inflammations and exhibit damaging properties through their proteases, contributing to the chronicity of the disease. Several recent studies identified useful biomarkers and targets for future therapeutics.

SUMMARY

The findings we review in this manuscript are of utmost importance in unraveling the complexity of CRS and provide us with the necessary knowledge for future clinical practices.

A Decade of Clinical Advances in Chronic Rhinosinusitis: 2012-2022

The field of chronic rhinosinusitis (CRS) is constantly evolving. In the past 10 years, key advancements in basic and translational research as well as clinical studies have improved our understanding and management of CRS. Notably, treatment options have expanded to include novel therapeutic drugs, devices, and surgical techniques. Assessments of patient symptoms and their impact on quality of life have become more standardized. Progress has also been made in both determining the true prevalence of CRS and recognizing comorbidities that can impact CRS severity. Practice guidelines have also shifted from expert opinion to more data-driven analyses. This review highlights major clinical advancements made in the field of CRS over the past 10 years as well as identifies current gaps in knowledge that can form the basis for new areas of study over the next decade.

Deep Learning-Based Prediction of Treatment Prognosis from Nasal Polyp Histology Slides

BACKGROUND

Histopathology of nasal polyps contains rich prognostic information, which is difficult to objectively extract. In the present study, we aimed to develop a prognostic indicator of patient outcomes by analyzing scanned conventional haematoxylin and eosin (H&E) -stained slides alone using deep learning.

METHODS

An interpretable supervised deep learning model was developed using 185 H&E-stained whole-slide images (WSIs) of nasal polyps, each from a patient randomly selected from the pool of 232 patients who underwent endoscopic sinus surgery at the First Affiliated Hospital of Sun Yat-sen University (internal cohort). We internally validated the model on a holdout dataset from the internal cohort (47 H&E-stained WSIs) and externally validated the model on 122 H&E-stained WSIs from the Seventh Affiliated Hospital of Sun Yat-sen University and the University of Hong Kong-Shenzhen Hospital (external cohort). A poor prognosis score (PPS) was established to evaluate patient outcomes, and then risk activation mapping was applied to visualize the histopathological features underlying PPS.

RESULTS

The model yielded a patient-level sensitivity of 79.5%, and specificity of 92.3%, with areas under the receiver operating characteristic curve of 0.943, on the multi-center external cohort. The predictive ability of PPS was superior to that of conventional tissue eosinophil number. Notably, eosinophil infiltration, goblet cell hyperplasia, glandular hyperplasia, squamous metaplasia, and fibrin deposition were identified as the main underlying features of PPS.

CONCLUSIONS

Our deep learning model is an effective method for decoding pathological images of nasal polyps, providing a valuable solution for disease prognosis prediction and precise patient treatment. This article is protected by copyright. All rights reserved.

Microvessel quantification by fully convolutional neural networks associated with type 2 inflammation in chronic rhinosinusitis

BACKGROUND

The pathogenesis of chronic rhinosinusitis (CRS) is still unclear, and little is known about angiogenesis in this disease. We utilized a fully convolutional network (FCN), which has been extensively used in image processing to study angiogenesis in CRS.

OBJECTIVE

To explore the tissue quantification of microvessels and their potential association with inflammation in CRS by using FCN to reflect the angiogenesis condition in CRS.

METHODS

For endotyping of CRS, tissue homogenates of 79 patients with CRS who had undergone functional endoscopic sinus surgery and 17 control subjects were analyzed for interferon gamma, transforming growth factor beta, interleukin (IL)-1β, IL-5, IL-6, IL-8, IL-10, IL-17, tumor necrosis factor alpha, eosinophilic cationic protein, immunoglobulin E, and Staphylococcus aureus-immunoglobulin E(SE-IgE). A total of 552 hematoxylin and eosin-stained images of 27 CRS tissue samples were used to develop an FCN, going through training, validation, and evaluation processes. An optimized FCN was applied to quantify the microvessels of tissue samples of all subjects. Correlation analysis between microvessel quantification with phenotype, endotype, clinical characteristics, and cytokine expression of CRS was carried out.

RESULTS

Quantification of microvessels in type 2 and non-type 2 CRS demonstrated considerable differences, with a higher expression in type 2 CRS. There was a strong negative correlation between the area ratio of microvessels with tissue tumor necrosis factor alpha and transforming growth factor beta levels and a mildly positive correlation with tissue IL-5 and eosinophilic cationic protein concentration.

CONCLUSION

FCN proved to facilitate the analysis of microvessels in airway tissue samples. This study elucidated the close association of angiogenesis with endotyping, suggesting that treatment aiming at antagonizing angiogenesis may assist to the therapy for the recrudescent and refractory CRS.

Role of Respiratory Epithelial Cells in Allergic Diseases

The airway epithelium provides the first line of defense to the surrounding environment. However, dysfunctions of this physical barrier are frequently observed in allergic diseases, which are tightly connected with pro- or anti-inflammatory processes. When the epithelial cells are confronted with allergens or pathogens, specific response mechanisms are set in motion, which in homeostasis, lead to the elimination of the invaders and leave permanent traces on the respiratory epithelium. However, allergens can also cause damage in the sensitized organism, which can be ascribed to the excessive immune reactions. The tight interaction of epithelial cells of the upper and lower airways with local and systemic immune cells can leave an imprint that may mirror the pathophysiology. The interaction with effector T cells, along with the macrophages, play an important role in this response, as reflected in the gene expression profiles (transcriptomes) of the epithelial cells, as well as in the secretory pattern (secretomes). Further, the storage of information from past exposures as memories within discrete cell types may allow a tissue to inform and fundamentally alter its future responses. Recently, several lines of evidence have highlighted the contributions from myeloid cells, lymphoid cells, stromal cells, mast cells, and epithelial cells to the emerging concepts of inflammatory memory and trained immunity.

Chinese Central Compartment Atopic Disease: The Clinical Characteristics and Cellular Endotypes Based on Whole-Slide Imaging

Purpose

Histopathologic characterizations of central compartment atopic disease (CCAD) by whole-slide imaging remains lacking. We aim to study clinical presentations and cellular endotyping diagnosis of Chinese CCAD using artificial intelligence (AI).

Methods

A total of 72 patients diagnosed with chronic rhinosinusitis with nasal polyps (CRSwNP) were enrolled. CCAD was defined by positive result of serology specific IgE, endoscopic and radiological findings. The aeroallergen sensitization status, endoscopic results, radiological findings, and symptoms were evaluated and compared between patients with CCAD (n=14), eosinophilic CRSwNP (ENP, n=32) and non-eosinophilic CRSwNP (NENP, n=26). The cellular endotypes including eosinophils, neutrophils, lymphocytes, and plasma cells were analyzed by the AI chronic rhinosinusitis evaluation platform 2.0.

Results

CCAD was most common in male (71.43%). The positive rate of aeroallergen in patients with CCAD is 100%, which is much higher than those in patients with ENP (40.63%) and NENP (23.08%). Allergic rhinitis incidence was found to be 57.14% in Chinese CCAD subjects, which is obviously higher when compared with those in patients with ENP (21.88%) or NENP (0.00%). The presence of asthma was not significantly different between groups. Chinese CCAD population demonstrated mild symptoms and lower endoscopic and radiological scores than those in patients with ENP and NENP. For cellular endotypes in CCAD subjects, the median of eosinophils, neutrophils, lymphocytes, and plasma cells was 26.55%, 0.49%, 60.85%, and 7.33%, respectively. The proportion of eosinophils in nasal tissue and peripheral blood mononuclear cells from the CCAD group is between the proportions in those patients with ENP and NENP.

Conclusion

Chinese CCAD was associated with aeroallergen sensitivity, and displayed an eosinophil-dominant inflammatory pattern. Thus, proper management with allergy control and topical steroids could be recommended for CCAD treatment.