Background: Odontogenic keratocysts (OKCs) are enigmatic developmental cysts that deserve special attention due to their heterogeneous appearance in histopathological characteristics and high recurrence rate. Despite several nomenclatures for classification, clinicians still confront challenges in its diagnosis and predicting its recurrence. This paper proposes an ensemble deep-learning-based prognostic and prediction algorithm, for the recurrence of sporadic odontogenic keratocysts, on hematoxylin and eosin stained pathological images of incisional biopsies before treatment. (2) Materials and Methods: In this study, we applied a deep-learning algorithm to an ensemble approach integrated with DenseNet-121, Inception-V3, and Inception-Resnet-V3 classifiers. Around 1660 hematoxylin and eosin stained pathologically annotated digital images of OKC-diagnosed (60) patients were supplied to train and predict recurrent OKCs. (3) Results: The presence of SEH (p = 0.004), an incomplete epithelial lining, (p = 0.023), and a corrugated surface (p = 0.049) were the most significant histological parameters distinguishing recurrent and non-recurrent OKCs. Amongst the classifiers, DenseNet-121 showed 93% accuracy in predicting recurrent OKCs. Furthermore, integrating and training the traditional ensemble model showed an accuracy of 95% and an AUC of 0.9872, with an execution time of 192.9 s. In comparison, our proposed model showed 97% accuracy with an execution time of 154.6 s. (4) Conclusions: Considering the outcome of our novel ensemble model, based on accuracy and execution time, the presented design could be embedded into a computer-aided design system for automation of risk stratification of odontogenic keratocysts.

Ensemble deep-learning-based prognostic and prediction for recurrence of sporadic odontogenic keratocysts on hematoxylin and eosin stained pathological images of incisional biopsies

Reda R.
Conceptualization
;
Testarelli L.
Ultimo
Conceptualization
2022

Abstract

Background: Odontogenic keratocysts (OKCs) are enigmatic developmental cysts that deserve special attention due to their heterogeneous appearance in histopathological characteristics and high recurrence rate. Despite several nomenclatures for classification, clinicians still confront challenges in its diagnosis and predicting its recurrence. This paper proposes an ensemble deep-learning-based prognostic and prediction algorithm, for the recurrence of sporadic odontogenic keratocysts, on hematoxylin and eosin stained pathological images of incisional biopsies before treatment. (2) Materials and Methods: In this study, we applied a deep-learning algorithm to an ensemble approach integrated with DenseNet-121, Inception-V3, and Inception-Resnet-V3 classifiers. Around 1660 hematoxylin and eosin stained pathologically annotated digital images of OKC-diagnosed (60) patients were supplied to train and predict recurrent OKCs. (3) Results: The presence of SEH (p = 0.004), an incomplete epithelial lining, (p = 0.023), and a corrugated surface (p = 0.049) were the most significant histological parameters distinguishing recurrent and non-recurrent OKCs. Amongst the classifiers, DenseNet-121 showed 93% accuracy in predicting recurrent OKCs. Furthermore, integrating and training the traditional ensemble model showed an accuracy of 95% and an AUC of 0.9872, with an execution time of 192.9 s. In comparison, our proposed model showed 97% accuracy with an execution time of 154.6 s. (4) Conclusions: Considering the outcome of our novel ensemble model, based on accuracy and execution time, the presented design could be embedded into a computer-aided design system for automation of risk stratification of odontogenic keratocysts.
File allegati a questo prodotto
File Dimensione Formato  
Rao_Ensemble_2022.pdf

accesso aperto

Tipologia: Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza: Creative commons
Dimensione 2.07 MB
Formato Adobe PDF
2.07 MB Adobe PDF Visualizza/Apri PDF

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11573/1654198
Citazioni
  • ???jsp.display-item.citation.pmc??? 0
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact