A comparative study on deep feature selection methods for skin lesion classification DOI Creative Commons

Farzad Golnoori,

Farsad Zamani Boroujeni,

Seyed Amirhassan Monadjemi

et al.

IET Image Processing, Journal Year: 2023, Volume and Issue: 18(4), P. 996 - 1013

Published: Nov. 27, 2023

Abstract Melanoma, a widespread and hazardous form of cancer, has prompted researchers to prioritize dermoscopic image‐based algorithms for classifying skin lesions. Recently, there been growing trend in using pre‐trained convolutional neural networks detecting However, the features extracted from these classifiers may include irrelevant elements, emphasizing importance implementing effective feature selection methods. Nevertheless, not comprehensive study on methods enhance performance lesion detection date. To identify most efficient methods, diverse set techniques, including filter, wrapper, embedded, dimensionality reduction, were applied images two well‐known datasets, namely ISIC 2017 2018. According results, models trained with chosen by wrapper techniques outperformed those filter embedded Achieving an accuracy 0.8333 F1‐Score 0.8291 dataset, 0.9324 0.9350 2018 classification obtained via GWO technique performed best.

Language: Английский

A new intelligent hybrid feature extraction model for automating cancer diagnosis: a focus on breast cancer DOI
Rasoul Rahmani, Shahin Akbarpour, Ali Farzan

et al.

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(5)

Published: March 24, 2025

Language: Английский

Citations

0
Esophageal Cancer Diagnosis with a Bilinear Pooling and Attention-Based Convolutional Neural Network DOI
Vikas Raina, Haewon Byeon, Manisha Bhende

et al.

Communications in computer and information science, Journal Year: 2025, Volume and Issue: unknown, P. 328 - 340

Published: Jan. 1, 2025

Language: Английский

Citations

0
AWFCNET: An Attention-Aware Deep Learning Network with Fusion Classifier for Breast Cancer Classification Using Enhanced Mammograms DOI
Renato R. Maaliw, Mukesh Soni,

Manuel P. Delos Santos

et al.

2022 IEEE World AI IoT Congress (AIIoT), Journal Year: 2023, Volume and Issue: unknown, P. 0736 - 0744

Published: June 7, 2023

Breast cancer remains a significant public health concern and leading cause of female mortality despite recent advances in healthcare. Experts agree that its early prognosis is key to survivability. In this research, we proposed deep learning architecture code-named AWFCNET. It comprised multiple segments preprocessing techniques (color shifting & image enhancement), feature based on ResNeXt-101 convolutional network as backbone with transfer attention-aware mechanisms, fusion classifier composed three recurrent neural networks. The generalization capability the pipeline produced 98.10% accuracy mammogram dataset using 10-fold cross-validation. Computational benchmarks revealed it surpassed existing state-of-the-art approaches provisions visual interpretability via gradient maps. Thus, our framework could complement physicians' expertise rapid dependable breast diagnoses.

Language: Английский

Citations

7
A computer-aided feature-based encryption model with concealed access structure for medical Internet of Things DOI Creative Commons

Sumit Vaidya,

Ashish Suri, Vishnu Batla

et al.

Decision Analytics Journal, Journal Year: 2023, Volume and Issue: 7, P. 100257 - 100257

Published: June 1, 2023

One of the Internet Things (IoT) security issues is secure sharing and granular management data access. This study recommends a feature-based encryption scheme with hidden access structure for medical IoT security. While establishing fine-grained control ciphertext data, system can guarantee clinical client privacy. First, it recommended to convert identity-based (IBE) into model (FBEM) using universal conversion technique that supports multi-valued attributes gates. IBE characteristics could be inherited by converted FBEM. The method then used change receiver anonymous FBEM concealed structure. construct scenario smart application. Theoretical analysis experimental findings reveal suggested provides advantages over prominent systems regarding computing efficiency, storage load, when disguised.

Language: Английский

Citations

6
Dimension Reduction and Classifier-Based Feature Selection for Oversampled Gene Expression Data and Cancer Classification DOI Open Access
Olutomilayo Olayemi Petinrin, Faisal Saeed, Naomie Salim

et al.

Processes, Journal Year: 2023, Volume and Issue: 11(7), P. 1940 - 1940

Published: June 27, 2023

Gene expression data are usually known for having a large number of features. Usually, some these features irrelevant and redundant. However, in cases, all features, despite being numerous, show high importance contribute to the analysis. In similar fashion, gene sometimes have limited instances with rate imbalance among classes. This can limit exposure classification model different categories, thereby influencing performance model. this study, we proposed cancer detection approach that utilized preprocessing techniques such as oversampling, feature selection, models. The study used SVMSMOTE oversampling six examined datasets. Further, selection using dimension reduction methods classifier-based ranking selection. We trained machine learning algorithms, repeated 5-fold cross-validation on microarray algorithms differed based technique used.

Language: Английский

Citations

5
Clustering-based spatial analysis (CluSA) framework through graph neural network for chronic kidney disease prediction using histopathology images DOI Creative Commons
Joonsang Lee, Elisa Warner, Salma Shaikhouni

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Aug. 5, 2023

Abstract Machine learning applied to digital pathology has been increasingly used assess kidney function and diagnose the underlying cause of chronic disease (CKD). We developed a novel computational framework, clustering-based spatial analysis (CluSA), that leverages unsupervised learn relationships between local visual patterns in tissue. This framework minimizes need for time-consuming impractical expert annotations. 107,471 histopathology images obtained from 172 biopsy cores were clustering deep model. To incorporate information over clustered image on sample, we spatially encoded with colors performed through graph neural network. A random forest classifier various groups features predict CKD. For predicting eGFR at biopsy, achieved sensitivity 0.97, specificity 0.90, accuracy 0.95. AUC was 0.96. changes one-year, 0.83, 0.85, 0.84. 0.85. study presents first based machine algorithms. Without annotation, CluSA can not only accurately classify degree one year, but also identify predictors renal prognosis.

Language: Английский

Citations

5
A comparative study on deep feature selection methods for skin lesion classification DOI Creative Commons

Farzad Golnoori,

Farsad Zamani Boroujeni,

Seyed Amirhassan Monadjemi

et al.

IET Image Processing, Journal Year: 2023, Volume and Issue: 18(4), P. 996 - 1013

Published: Nov. 27, 2023

Abstract Melanoma, a widespread and hazardous form of cancer, has prompted researchers to prioritize dermoscopic image‐based algorithms for classifying skin lesions. Recently, there been growing trend in using pre‐trained convolutional neural networks detecting However, the features extracted from these classifiers may include irrelevant elements, emphasizing importance implementing effective feature selection methods. Nevertheless, not comprehensive study on methods enhance performance lesion detection date. To identify most efficient methods, diverse set techniques, including filter, wrapper, embedded, dimensionality reduction, were applied images two well‐known datasets, namely ISIC 2017 2018. According results, models trained with chosen by wrapper techniques outperformed those filter embedded Achieving an accuracy 0.8333 F1‐Score 0.8291 dataset, 0.9324 0.9350 2018 classification obtained via GWO technique performed best.

Language: Английский

Citations

2