CervixNet: A Reward-Based Weighted Ensemble Framework for Cervical Cancer Classification

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F24%3A50021552" target="_blank" >RIV/62690094:18450/24:50021552 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1007/978-981-97-2611-0_20" target="_blank" >http://dx.doi.org/10.1007/978-981-97-2611-0_20</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-981-97-2611-0_20" target="_blank" >10.1007/978-981-97-2611-0_20</a>

Jazyk výsledku

angličtina

Název v původním jazyce

CervixNet: A Reward-Based Weighted Ensemble Framework for Cervical Cancer Classification

Popis výsledku v původním jazyce

Cervical cancer is one of the most common causes of death among women worldwide. However, this fatal disease can be treated, and the mortality rate can be decreased if detected at an early stage. The Papanicolaou test is the gold standard for screening cervical cancer patients. However, the process of manual inspection is tedious and subject to manual errors. Thus, computer-based automatic screening is considered a viable alternative. To this end, in the present work, we have developed a novel system for classifying cervical cancer images as either normal or malignant. We have used the Herlev dataset in this study, which is the standard Pap smear image benchmark dataset. The proposed framework uses multiple base deep learning frameworks, namely Vision Transformer, Xception, VGG-19, VGG-16, and ResNet-101, for classification. Further, we have introduced a novel reward-based weighing technique to decide the weights of individual classifiers, which are, in turn, used to make decisions about the final class label of an input image using the weighted average technique. The proposed framework achieves an overall accuracy, precision, recall, F1-score, and AUC of 96%, 94%, 92%, 93%, and 94%, respectively, for the binary classification task that is normal vs. malignant. An elaborate study of the performance achieved by our proposed framework on the Herlev dataset shows it to be both robust and effective for the Pap smear image classification task.

Název v anglickém jazyce

CervixNet: A Reward-Based Weighted Ensemble Framework for Cervical Cancer Classification

Popis výsledku anglicky

Cervical cancer is one of the most common causes of death among women worldwide. However, this fatal disease can be treated, and the mortality rate can be decreased if detected at an early stage. The Papanicolaou test is the gold standard for screening cervical cancer patients. However, the process of manual inspection is tedious and subject to manual errors. Thus, computer-based automatic screening is considered a viable alternative. To this end, in the present work, we have developed a novel system for classifying cervical cancer images as either normal or malignant. We have used the Herlev dataset in this study, which is the standard Pap smear image benchmark dataset. The proposed framework uses multiple base deep learning frameworks, namely Vision Transformer, Xception, VGG-19, VGG-16, and ResNet-101, for classification. Further, we have introduced a novel reward-based weighing technique to decide the weights of individual classifiers, which are, in turn, used to make decisions about the final class label of an input image using the weighted average technique. The proposed framework achieves an overall accuracy, precision, recall, F1-score, and AUC of 96%, 94%, 92%, 93%, and 94%, respectively, for the binary classification task that is normal vs. malignant. An elaborate study of the performance achieved by our proposed framework on the Herlev dataset shows it to be both robust and effective for the Pap smear image classification task.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of 4th International Conference on Frontiers in Computing and Systems

ISBN

978-981-9726-10-3

ISSN

2367-3370

e-ISSN

2367-3389

Počet stran výsledku

13

Strana od-do

293-305

Název nakladatele

Springer Singapore

Místo vydání

Singapore

Místo konání akce

Mandi, India

Datum konání akce

16. 10. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—