Multiclass Skin Cancer Classification Using Ensemble of Fine-Tuned Deep Learning Models

Kausar, Nabeela; Hameed, Abdul; Sattar, Mohsin; Ashraf, Ramiza; Imran, Ali Shariq; Abidin, Muhammad Zain ul; Ali, Ammara

doi:10.3390/app112210593

Cited by 37 publications

(18 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is known from the literature that the accuracy of classi cation declines with the number of classes. The accuracy of the earlier studies reported in [12,13,14,15] is lower than that of the ensemble models that have been suggested. With minimal pre-processing, our ensemble models have outperformed contemporary deep learning-based models as well as dermatologists in the multiclass classi cation of skin cancer.…”

Section: Classi Cation Reportcontrasting

confidence: 66%

“…The proposed ensemble model's performance is compared to the most recent deep learning-based models proposed in [12,13,14,15] in Figures 12,15,and 18. The accuracy of the majority voting, weighted averaging, and weighted majority ensemble models is seen to be substantially greater than that of the models provided in [12,13,14,15]. It is known from the literature that the accuracy of classi cation declines with the number of classes.…”

Section: Classi Cation Reportmentioning

confidence: 99%

See 1 more Smart Citation

Multiclass Skin Cancer Classification using Ensembled Deep Learning Model

Prakash

Acharya

Hiriyannaiah

et al. 2022

Preprint

View full text Add to dashboard Cite

Skin cancer is a major health concern, with 125,000 new melanoma cases diagnosed each year. Every year, approximately 3 million non-melanoma cases are diagnosed worldwide. Initially, skin lesions were diagnosed visually using dermoscopic analysis, clinical screening, and a variety of other methods. It has been observed that inexperienced dermatologists can reduce the diagnostic accuracy of skin lesions. Early detection of skin cancer has the potential to reduce mortality. Previous research has shown that deep learning outperforms human experts in a variety of computer vision tasks. This paper proposes to develop an ensemble deep learning model to detect skin cancer from lesion images.In this analytical investigation, 10,000 photos from the HAM10000 dermoscopy image database were used, comprising lesions with and without melanoma.The photos were divided into different skin cancer classifications using a number of deep convolutional neural networks.VGG16, VGG19, and ResNet were used as the pre-trained models in an ensemble deep learning approach. The area under the receiver operating characteristic (ROC) curve for the suggested model was 0.91. A classification accuracy of 86%, sensitivity of 81%, and specificity of 88% were attained using a confidence score threshold of 0.5. Detecting skin cancer, including melanoma and non-melanoma tumors, has a high potential according to the results of deep learning. Dermatologists can identify skin cancer with the use of the indicated method.As a result, it might hasten cancer detection, which is essential for efficient therapy. The suggested model automatically extracts useful features from the input raw image for categorization.It does away with difficult lesion segmentation and feature extraction processes as a result.

show abstract

Section: Classi Cation Reportcontrasting

confidence: 66%

Section: Classi Cation Reportmentioning

confidence: 99%

Multiclass Skin Cancer Classification using Ensembled Deep Learning Model

Prakash

Acharya

Hiriyannaiah

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The recent ensemble model in [30,31] achieved accuracy higher than the proposed ensemble models because they ensemble more CNN models that is lead to complexity in the process and used an extra dataset. The recent ensemble model in [33] achieved accuracy higher than the proposed ensemble models because they implemented their experiments on part of dataset by selecting randomly images from each class. The recent ensemble model in [34] achieved accuracy higher than the proposed ensemble models because they ensemble five models to diagnose seven classes of skin lesions types by only using seven classes from ISIC 2019 like HAM dataset and the classification accuracy decrease when the number of diagnostic classes increases.…”

Section: Resultsmentioning

confidence: 97%

“…Table IV shows the comparison between the proposed ensemble models and the recent ensemble models in [29,30,31,32,33,34,35] Although the recent ensemble models used the normal ensemble algorithms but some of them achieved higher accuracy than proposed ensemble learning model because using extra dataset and ensemble with more classification deep learner models lead to increasing the accuracy result. The recent ensemble model in [30,31] achieved accuracy higher than the proposed ensemble models because they ensemble more CNN models that is lead to complexity in the process and used an extra dataset.…”

Section: Resultsmentioning

confidence: 99%

An Enhanced Technique for Skin Lesion Diagnosis using Dermoscopic Images

Mosa

Afifi

Amin

2022

IJCI. International Journal of Computers and Information

View full text Add to dashboard Cite

There are many types of skin cancer, the most harmful among them is melanoma. Skin cancer occurs through the abnormal growth of the body cells, which may be caused by continuous exposure to ultraviolet rays resulting from the sun. Early diagnosis of skin cancer is essential as it can reduce the burden, make the treatment more effective and save the patient life. In this work, therefore, we develop an enhanced ensemble approach to improve the classification accuracy of eight types of skin cancer. Three pretrained Convolutional Neural Network (CNN) models, namely ResNet18, DensNet121, and Inception v4, are used as a base for this ensemble. Firstly, we fine-tune each pre-trained CNN model separately on an augmented dataset. Afterwards, the prediction probabilities obtained from all the base learners are combined as a new feature vector for each case. These features are used to build another classifier that employs the outputs of all the networks used in the previous stage. Thus, this ensemble learns how to weigh the output of each base learner without any user interaction. Support vector machine (SVM), and random forest (RF) classifiers are utilized in this stage and compared to the average ensemble which gives the same importance to all base learners.The pre-trained CNN models were fine-tuned and evaluated using 17731 and 3800 images from different types of skin cancer, respectively. The ResNet18, DensNet121, and InceptionV4 achieved individual accuracy of 79.5%, 81.2%, and 82.6%, respectively. The proposed ensemble method using SVM, and RF classifiers gives the best accuracy result with 85% for the SVM classifier and 86.2% for the Random Forest classifier. These results show that the proposed ensemble method using SVM, and RF classifiers is superior to the single classifiers of ResNet18, DensNet121 and InceptionV4 as well as the commonly used average ensemble which achieved an accuracy of 83.2%.

show abstract

“…With the help of this technique, they claim that their model accuracy was enhanced from 66% to 92%. In another study Kausar et al [34] used some fine-tuning techniques to improve state-of-art deep learning image classification models. They achieved an accuracy of 72%, 91%, 91.4%, 91.7%, and 91.8% for ResNet, InceptionV3, DenseNet, InceptionResNetV2, and VGG-19, respectively.…”

Section: Introductionmentioning

confidence: 99%

SBXception: A Shallower and Broader Xception Architecture for Efficient Classification of Skin Lesions

Mehmood

Gulzar

Ilyas

et al. 2023

Cancers

View full text Add to dashboard Cite

Skin cancer is a major public health concern around the world. Skin cancer identification is critical for effective treatment and improved results. Deep learning models have shown considerable promise in assisting dermatologists in skin cancer diagnosis. This study proposes SBXception: a shallower and broader variant of the Xception network. It uses Xception as the base model for skin cancer classification and increases its performance by reducing the depth and expanding the breadth of the architecture. We used the HAM10000 dataset, which contains 10,015 dermatoscopic images of skin lesions classified into seven categories, for training and testing the proposed model. Using the HAM10000 dataset, we fine-tuned the new model and reached an accuracy of 96.97% on a holdout test set. SBXception also achieved significant performance enhancement with 54.27% fewer training parameters and reduced training time compared to the base model. Our findings show that reducing and expanding the Xception model architecture can greatly improve its performance in skin cancer categorization.

show abstract

Multiclass Skin Cancer Classification Using Ensemble of Fine-Tuned Deep Learning Models

Cited by 37 publications

References 56 publications

Multiclass Skin Cancer Classification using Ensembled Deep Learning Model

Multiclass Skin Cancer Classification using Ensembled Deep Learning Model

An Enhanced Technique for Skin Lesion Diagnosis using Dermoscopic Images

SBXception: A Shallower and Broader Xception Architecture for Efficient Classification of Skin Lesions

Contact Info

Product

Resources

About