Automated Skin Cancer Detection System Using Deep Transfer Learning DOI

S H Shruthishree

Published: Feb. 16, 2024

One of the most prevalent cancers, both melanoma and non-melanoma, causes hundreds thousands deaths globally each year. Skin cell growth that isn't normal is how it shows up. Recovery chances are significantly increased by early diagnosis. Furthermore, might reduce need for or use chemical, radiographic, surgical therapies altogether. A dermatoscope used in traditional method visual inspection a dermatologist primary care physician order to detect skin-related diseases. Patients who exhibit signs skin cancer referred biopsy histopathological examination confirm diagnosis determine appropriate course treatment. Recent developments deep convolutional neural networks (CNNs) have led automated classification with excellent performance accuracy comparable dermatologists. These advancements haven't, however, yet produced widely clinically reliable identification cancer. As result, medical expenses can be decreased. Dermoscopy, which examines general size, shape, color characteristics lesions, first step Suspected lesions then undergo additional sampling laboratory testing confirmation. Because learning artificial intelligence has become more popular, image-based advanced recent years.

Language: Английский

Vision Transformers for Low-Quality Histopathological Images: A Case Study on Squamous Cell Carcinoma Margin Classification DOI Creative Commons

Saeran Park,

Gelan Ayana, Beshatu Debela Wako

et al.

Diagnostics, Journal Year: 2025, Volume and Issue: 15(3), P. 260 - 260

Published: Jan. 23, 2025

Background/Objectives: Squamous cell carcinoma (SCC), a prevalent form of skin cancer, presents diagnostic challenges, particularly in resource-limited settings with low-quality imaging infrastructure. The accurate classification SCC margins is essential to guide effective surgical interventions and reduce recurrence rates. This study proposes vision transformer (ViT)-based model improve margin by addressing the limitations convolutional neural networks (CNNs) analyzing histopathological images. Methods: introduced transfer learning approach using ViT architecture customized additional flattening, batch normalization, dense layers enhance its capability for classification. A performance evaluation was conducted machine metrics averaged over five-fold cross-validation comparisons were made leading CNN models. Ablation studies have explored effects architectural configuration on performance. Results: ViT-based achieved superior 0.928 ± 0.027 accuracy 0.927 0.028 AUC, surpassing highest performing model, InceptionV3 (accuracy: 0.86 0.049; AUC: 0.837 0.029), demonstrating robustness reinforced importance tailored configurations enhancing Conclusions: underscores transformative potential ViTs analysis, especially settings. By reducing dependence high-quality specialized expertise, it scalable solution global cancer diagnostics. Future research should prioritize optimizing such environments broadening their clinical applications.

Language: Английский

Citations

0
The Use of Machine Learning to Support the Diagnosis of Oral Alterations DOI Creative Commons
Rosana Leal do Prado, Juliane Avansini Marsicano, Amanda Keren Frois

et al.

Pesquisa Brasileira em Odontopediatria e Clínica Integrada, Journal Year: 2025, Volume and Issue: 25

Published: Jan. 1, 2025

ABSTRACT Objective: To verify the accuracy of deep learning models in detecting cellular alterations histological images oral mucosa. Material and Methods: The study compares three convolutional neural network (CNN) architectures for classifying images: EfficientNet-B3, MobileNet-V2, VGG16. Efficient focused on computer vision, each has specific advantages. A Kaggle database with 5192 was used, divided into training (70%), validation (15%), test (15%) sets. CNNs were implemented using Keras library, trained pre-trained ImageNet weights, evaluated AUC metrics. Results: findings indicate that EfficientNet-B3 achieved lowest losses at epoch 30, highest stability during training. Evaluation metrics showed 98% 99% sensitivity squamous cell carcinoma (OSCC) images, outperforming MobileNet-V2 97% 96% sensitivity, while VGG16 reached 94% 93% OSCC images. All exhibited high specificity differentiating between normal as demonstrated by ROC curves. had (0.982), followed (AUC=0.967) (AUC=0.937). These underscore effectiveness accurately Conclusion: Our reveals superior performance CNNs, particularly OSCC.

Language: Английский

Citations

0
Deep structured learning with vision intelligence for oral carcinoma lesion segmentation and classification using medical imaging DOI Creative Commons
Ahmad A. Alzahrani, Jamal Alsamri, Mashael Maashi

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 24, 2025

Abstract Oral carcinoma (OC) is a toxic illness among the most general malignant cancers globally, and it has developed gradually significant public health concern in emerging low-to-middle-income states. Late diagnosis, high incidence, inadequate treatment strategies remain substantial challenges. Analysis at an initial phase for good treatment, prediction, existence. Despite current growth perception of molecular devices, late analysis methods near precision medicine OC patients challenge. A machine learning (ML) model was employed to improve early detection medicine, aiming reduce cancer-specific mortality disease progression. Recent advancements this approach have significantly enhanced extraction diagnosis critical information from medical images. This paper presents Deep Structured Learning with Vision Intelligence Carcinoma Lesion Segmentation Classification (DSLVI-OCLSC) imaging. Using imaging, DSLVI-OCLSC aims enhance OC’s classification recognition outcomes. To accomplish this, utilizes wiener filtering (WF) as pre-processing technique eliminate noise. In addition, ShuffleNetV2 method used group higher-level deep features input image. The convolutional bidirectional long short-term memory network multi-head attention mechanism (MA-CNN‐BiLSTM) utilized oral identification. Moreover, Unet3 + segment abnormal regions classified Finally, sine cosine algorithm (SCA) hyperparameter-tune DL model. wide range simulations implemented ensure performance under images dataset. experimental portrayed superior accuracy value 98.47% over recent approaches.

Language: Английский

Citations

0
Deeppatchnet: A Deep Learning Model for Enhanced Screening and Diagnosis of Oral Cancer DOI
Idriss Tafala,

fatima-ezzahraa ben-bouazza,

Aymane Edder

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Ultrasensitive detection of PrPC in human serum using label-free electrochemical biosensor based on α-iron trioxide/ferriferrous oxide magnetic nanocomposites DOI

Haoda Zhang,

Zhixiang Lv, Hexiao Zhang

et al.

Microchemical Journal, Journal Year: 2025, Volume and Issue: unknown, P. 113369 - 113369

Published: March 1, 2025

Language: Английский

Citations

0
HistoMoCo: Momentum Contrastive Learning Pre-Training on Unlabeled Histopathological Images for Oral Squamous Cell Carcinoma Detection DOI Open Access
Weibin Liao, Yifan He, Bowen Jiang

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(7), P. 1252 - 1252

Published: March 22, 2025

The early detection and intervention of oral squamous cell carcinoma (OSCC) using histopathological images are crucial for improving patient outcomes. current literature identifying OSCC predominantly relies on models pre-trained ImageNet to minimize the need manual data annotations in model fine-tuning. However, a significant divergence exists between visual domains natural images, potentially limiting representation transferability these models. Inspired by recent self-supervised research, this work, we propose HistoMoCo, an adaptation Momentum Contrastive Learning (MoCo), designed generate with enhanced image representations initializations images. Specifically, HistoMoCo aggregates 102,228 leverages structure features unique histological data, allowing more robust feature extraction subsequent downstream We perform tasks evaluate two real-world datasets, including NDB-UFES Oral Histopathology datasets. Experimental results demonstrate that consistently outperforms traditional ImageNet-based pre-training, yielding stable accurate performance detection, achieving AUROC up 99.4% dataset 94.8% dataset. Furthermore, dataset, pre-training solution achieves 89.32% 40% training whereas reaches 89.58% only 10% data. addresses issue domain state-of-the-art More importantly, significantly reduces reliance release our code parameters further research histopathology or tasks.

Language: Английский

Citations

0
A robust transfer learning approach with histopathological images for lung and colon cancer detection using EfficientNetB3 DOI Creative Commons
Raquel Ochoa-Ornelas, Alberto Gudiño-Ochoa, J A García-Rodríguez

et al.

Healthcare Analytics, Journal Year: 2025, Volume and Issue: unknown, P. 100391 - 100391

Published: April 1, 2025

Language: Английский

Citations

0
Enhanced Layer Extraction for Efficient Plant Disease Classification using Efficientnet B1 DOI
Arastu Thakur, Awalendra K. Thakur, V Vivek

et al.

SN Computer Science, Journal Year: 2025, Volume and Issue: 6(4)

Published: April 11, 2025

Language: Английский

Citations

0
Deep Learning-Based Automated Detection of Oral Leukoplakia in Clinical Imaging DOI Open Access
Duo Li, Xiangjian Wang, Jingwen Liu

et al.

Cureus, Journal Year: 2025, Volume and Issue: unknown

Published: May 2, 2025

Language: Английский

Citations

0
Technology-Driven Financial Risk Management: Exploring the Benefits of Machine Learning for Non-Profit Organizations DOI Creative Commons
Hao Huang

Systems, Journal Year: 2024, Volume and Issue: 12(10), P. 416 - 416

Published: Oct. 8, 2024

This study explores how machine learning can optimize financial risk management for non-profit organizations by evaluating various algorithms aimed at mitigating loan default risks. The findings indicate that ensemble models, such as random forest and LightGBM, significantly improve prediction accuracy, thereby enabling non-profits to better manage risk. In the context of 2008 subprime mortgage crisis, which underscored volatility markets, this research assesses a range risks—credit, operational, liquidity, market risks—while exploring both traditional advanced techniques, with particular focus on stacking fusion enhance model performance. Emphasizing importance privacy adaptive methods, advocates interdisciplinary approaches overcome limitations stress testing, data analysis rule formulation, regulatory collaboration. underscores learning’s crucial role in control calls authorities reassess existing frameworks accommodate evolving Additionally, it highlights need accurate type identification potential strengthen amid uncertainty, promoting efforts address broader issues like environmental sustainability economic development.

Language: Английский

Citations

3