Enhanced skin cancer diagnosis: a deep feature extraction-based framework for the multi-classification of skin cancer utilizing dermoscopy images DOI Creative Commons
Hadeel Alharbi, Gabriel Avelino Sampedro, Roben A. Juanatas

и другие.

Frontiers in Medicine, Год журнала: 2024, Номер 11

Опубликована: Ноя. 13, 2024

Skin cancer is one of the most common, deadly, and widespread cancers worldwide. Early detection skin can lead to reduced death rates. A dermatologist or primary care physician use a dermatoscope inspect patient diagnose disorders visually. essential, in order confirm diagnosis determine appropriate course therapy, patients should undergo biopsy histological evaluation. Significant advancements have been made recently as accuracy categorization by automated deep learning systems matches that dermatologists. Though progress has made, there still lack widely accepted, clinically reliable method for diagnosing cancer. This article presented four variants Convolutional Neural Network (CNN) model (i.e., original CNN, no batch normalization few filters strided CNN) classification prediction lesion images with aim helping physicians their diagnosis. Further, it presents hybrid models CNN-Support Vector Machine (CNNSVM), CNN-Random Forest (CNNRF), CNN-Logistic Regression (CNNLR), using grid search best parameters. Exploratory Data Analysis (EDA) random oversampling are performed normalize balance data. The CNN (original strided, CNNSVM) obtained an rate 98%. In contrast, CNNRF CNNLR 99% on HAM10000 dataset 10,015 dermoscopic images. encouraging outcomes demonstrate effectiveness proposed show improving performance requires including patient's metadata image.

Язык: Английский

Assessing the Efficacy of the Spectrum-Aided Vision Enhancer (SAVE) to Detect Acral Lentiginous Melanoma, Melanoma In Situ, Nodular Melanoma, and Superficial Spreading Melanoma: Part II DOI Creative Commons
Teng‐Li Lin, Riya Karmakar, Arvind Mukundan

и другие.

Diagnostics, Год журнала: 2025, Номер 15(6), С. 714 - 714

Опубликована: Март 13, 2025

Background: Melanoma, a highly aggressive form of skin cancer, necessitates early detection to significantly improve survival rates. Traditional diagnostic techniques, such as white-light imaging (WLI), are effective but often struggle differentiate between melanoma subtypes in their stages. Methods: The emergence the Spectrum-Aided Vison Enhancer (SAVE) offers promising alternative by utilizing specific wavelength bands enhance visual contrast lesions. This technique facilitates greater differentiation malignant and benign tissues, particularly challenging cases. In this study, efficacy SAVE is evaluated detecting including acral lentiginous (ALM), situ (MIS), nodular (NM), superficial spreading (SSM) compared WLI. Results: findings demonstrated that consistently outperforms WLI across various key metrics, precision, recall, F1-scorw, mAP, making it more reliable tool for using four different machine learning methods YOLOv10, Faster RCNN, Scaled YOLOv4, YOLOv7. Conclusions: ability capture subtle spectral differences clinicians new avenue improving accuracy patient outcomes.

Язык: Английский

Процитировано

0
Skin cancer severity analysis and prediction framework based on deep learning DOI

Runhe Huang,

Haoran Wu

Опубликована: Окт. 25, 2024

Язык: Английский

Процитировано

1
Enhanced skin cancer diagnosis: a deep feature extraction-based framework for the multi-classification of skin cancer utilizing dermoscopy images DOI Creative Commons
Hadeel Alharbi, Gabriel Avelino Sampedro, Roben A. Juanatas

и другие.

Frontiers in Medicine, Год журнала: 2024, Номер 11

Опубликована: Ноя. 13, 2024

Skin cancer is one of the most common, deadly, and widespread cancers worldwide. Early detection skin can lead to reduced death rates. A dermatologist or primary care physician use a dermatoscope inspect patient diagnose disorders visually. essential, in order confirm diagnosis determine appropriate course therapy, patients should undergo biopsy histological evaluation. Significant advancements have been made recently as accuracy categorization by automated deep learning systems matches that dermatologists. Though progress has made, there still lack widely accepted, clinically reliable method for diagnosing cancer. This article presented four variants Convolutional Neural Network (CNN) model (i.e., original CNN, no batch normalization few filters strided CNN) classification prediction lesion images with aim helping physicians their diagnosis. Further, it presents hybrid models CNN-Support Vector Machine (CNNSVM), CNN-Random Forest (CNNRF), CNN-Logistic Regression (CNNLR), using grid search best parameters. Exploratory Data Analysis (EDA) random oversampling are performed normalize balance data. The CNN (original strided, CNNSVM) obtained an rate 98%. In contrast, CNNRF CNNLR 99% on HAM10000 dataset 10,015 dermoscopic images. encouraging outcomes demonstrate effectiveness proposed show improving performance requires including patient's metadata image.

Язык: Английский

Процитировано

0