Colonoscopy polyp classification via enhanced scattering wavelet Convolutional Neural Network DOI Creative Commons
Jun Tan,

Jiamin Yuan,

Xiaoyong Fu

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(10), P. e0302800 - e0302800

Published: Oct. 11, 2024

Among the most common cancers, colorectal cancer (CRC) has a high death rate. The best way to screen for is with colonoscopy, which been shown lower risk of disease. As result, Computer-aided polyp classification technique applied identify cancer. But visually categorizing polyps difficult since different have lighting conditions. Different from previous works, this article presents Enhanced Scattering Wavelet Convolutional Neural Network (ESWCNN), that combines (CNN) and Transform (SWT) improve performance. This method concatenates simultaneously learnable image filters wavelet on each input channel. scattering can extract spectral features various scales orientations, while capture spatial may miss. A network architecture ESWCNN designed based these principles trained tested using colonoscopy datasets (two public one private dataset). An n-fold cross-validation experiment was conducted three classes (adenoma, hyperplastic, serrated) achieving accuracy 96.4%, 94.8% in two-class (positive negative). In three-class classification, correct rates 96.2% adenomas, 98.71% hyperplastic polyps, 97.9% serrated were achieved. proposed reached an average sensitivity 96.7% 93.1% specificity. Furthermore, we compare performance our model state-of-the-art general models commonly used CNNs. Six end-to-end CNNs 2 dataset video sequences. experimental results demonstrate effectively classify higher efficacy compared CNN models. These findings provide guidance future research classification.

Language: Английский

Advanced deep learning algorithms in oral cancer detection: Techniques and applications DOI

Dipali Wankhade,

Chitra Dhawale,

Mrunal Meshram

et al.

Journal of Environmental Science and Health Part C, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 26

Published: Jan. 17, 2025

As the 16th most common cancer globally, oral yearly accounts for some 355,000 new cases. This study underlines that an early diagnosis can improve prognosis and cut down on mortality. It discloses a multifaceted approach to detection of cancer, including clinical examination, biopsies, imaging techniques, incorporation artificial intelligence deep learning methods. is distinctive in it provides thorough analysis recent AI-based methods detecting models machine algorithms use convolutional neural networks. By improving precision effectiveness cell detection, these eventually make therapy possible. also discusses importance techniques image pre-processing segmentation quality feature extraction, essential component accurate diagnosis. These have shown promising results, with classification accuracies reaching up 97.66% models. Integrating conventional cutting-edge AI technologies, this seeks advance thus enhancing patient outcomes cutting burden disease imposing healthcare systems.

Language: Английский

Citations

1
Detection of oral squamous cell carcinoma cancer using AlexNet on histopathological images DOI Creative Commons
Nigus Wereta Asnake, Aleka Melese Ayalew,

Abebaw Agegne Engda

et al.

Deleted Journal, Journal Year: 2025, Volume and Issue: 7(3)

Published: Feb. 19, 2025

The increase of oral squamous cell carcinoma (OSCC) in many countries is primarily linked to its high mortality rate and poor forecast. diagnosis patients with OSCC usually made by a pathologist who has utilized decades useful data from tissue biopsy samples. Human error increases while trying identify the cells via hand photographs microscope Previous studies have applied convolutional neural networks (CNNs) pre-trained models detect diseases improve accuracy. However, this approach outcomes number false positives negatives, which may lead inaccurate diagnoses. To cancer histopathological images, we developed AlexNet deep learning model. We used image preprocessing methods enhance quality histopathology images utilizing bilateral filtering color normalization histogram enhancement. Additionally, AlexNet, MobileNetV3, InceptionV3 for feature extraction, as well XGBoost classifier. emphasize reliability our findings, expanded on evaluation metrics (accuracy, precision, recall, F1 score) achieved proposed model, particularly accuracy 99%, precision 98.5%, recall score highlighting model's strength detection. These findings highlight effectiveness transfer detecting medical providing significant advancements early detection making valuable contribution field oncology.

Language: Английский

Citations

0
Artificial intelligence based classification and prediction of medical imaging using a novel framework of inverted and self-attention deep neural network architecture DOI Creative Commons
Junaid Aftab, Muhammad Attique Khan, Sobia Arshad

et al.

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

Published: March 13, 2025

Classifying medical images is essential in computer-aided diagnosis (CAD). Although the recent success of deep learning classification tasks has proven advantages over traditional feature extraction techniques, it remains challenging due to inter and intra-class similarity caused by diversity imaging modalities (i.e., dermoscopy, mammography, wireless capsule endoscopy, CT). In this work, we proposed a novel deep-learning framework for classifying several modalities. training phase models, data augmentation performed at first stage on all selected datasets. After that, two custom architectures were introduced, called Inverted Residual Convolutional Neural Network (IRCNN) Self Attention CNN (SACNN). Both models are trained augmented datasets with manual hyperparameter selection. Each dataset's testing used extract features during stage. The extracted fused using modified serial fusion strong correlation approach. An optimization algorithm- slap swarm controlled standard Error mean (SScSEM) been employed, best that passed shallow wide neural network (SWNN) classifier final have selected. GradCAM, an explainable artificial intelligence (XAI) approach, analyzes models. architecture was tested five publically available different obtained improved accuracy 98.6 (INBreast), 95.3 (KVASIR), 94.3 (ISIC2018), 95.0 (Lung Cancer), 98.8% (Oral respectively. A detailed comparison conducted based precision accuracy, showing performs better. implemented GitHub ( https://github.com/ComputerVisionLabPMU/ScientificImagingPaper.git ).

Language: Английский

Citations

0
BFuse-Net: Bonferroni Mean Operator-Aided Fusion of Neural Networks for Medical Image Classification DOI

Triyas Ghosh,

Soham Chakraborty,

Dmitrii Kaplun

et al.

Lecture notes in electrical engineering, Journal Year: 2025, Volume and Issue: unknown, P. 100 - 109

Published: Jan. 1, 2025

Language: Английский

Citations

0
Classification of single-cell routine Pap smear images based on deep learning algorithms DOI Creative Commons

Guanhao Zhang

Theoretical and Natural Science, Journal Year: 2024, Volume and Issue: 50(1), P. 45 - 51

Published: Aug. 27, 2024

In this study, we classified single-cell routine Pap smear images by applying deep learning algorithms such as AlexNet, VggNet, GoogleNet and MobileNet compared their classification effects. The results show that the loss of all four models on both training test sets shows a trend gradually decreasing stabilising. Specifically, AlexNet decreases from 0.637 to 0.212, VggNet 0.777 0.278, 1.77 0.31, 0.809 0.267. At same time, exhibits highest maximum average accuracies which reached 93.9% 88.3%, respectively, followed model with 92.9% 88.0%, 92% 90.1% 86.7%. superior in task, provides strong support for its potential application images. These findings are great significance further exploring field medical imaging provide useful reference future related research.

Language: Английский

Citations

1
Demon Registration for 2D Empirical Wavelet Transforms DOI Creative Commons
Charles-Gérard Lucas, Jérôme Gilles

Foundations, Journal Year: 2024, Volume and Issue: 4(4), P. 690 - 703

Published: Dec. 3, 2024

The empirical wavelet transform is a fully adaptive time-scale representation that has been widely used in the last decade. Inspired by mode decomposition, it consists of filter banks based on harmonic supports. Recently, generalized to build from any generating function using mappings. In practice, supports can have low-constrained shape 2D, leading numerical difficulties estimate mappings adapted construction filters. This work aims propose an efficient scheme compute coefficients demons registration algorithm. Results show proposed approach robust, accurate, and continuous filters permitting reconstruction with low signal-to-noise ratio. An application for texture segmentation scanning tunneling microscope images also presented.

Language: Английский

Citations

1
Colonoscopy polyp classification via enhanced scattering wavelet convolutional neural network DOI Creative Commons
Jun Tan,

Jiamin Yuan,

Xiaoyong Fu

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 19, 2024

Abstract Among the most common cancers, colorectal cancer (CRC) has a high death rate. The best way to screen for is with colonoscopy, which been shown lower risk of disease. As result, Computer-aided polyp classification technique applied identify cancer. But visually categorizing polyps difficult since different have lighting conditions. Different from previous works, this article presents Enhanced Scattering Wavelet Convolutional Neural Network (ESWCNN), that combines (CNN) and Transform (SWT) improve performance. This method concatenates simultaneously learnable image filters wavelet on each input channel. scattering can extract spectral features various scales orientations, while capture spatial may miss. A network architecture ESWCNN designed based these principles trained tested using colonoscopy datasets (two public one private dataset). An n-fold cross-validation experiment was conducted three classes (adenoma, hyperplastic, serrated) achieving accuracy 96.4%, 94.8% in two-class (positive negative). In three-class classification, correct rates 96.2% adenomas, 98.71% hyperplastic polyps, 97.9% serrated were achieved. proposed reached an average sensitivity 96.7% 93.1% specificity. Furthermore, we compare performance our model state-of-the-art general models commonly used CNNs. Six end-to-end CNNs 2 dataset video sequences. experimental results demonstrate effectively classify higher efficacy compared CNN models. These findings provide guidance future research classification.

Language: Английский

Citations

0
Empirical mode decomposition and Hessian LLE in Fluorescence spectral signal analysis for Cervical cancer detection DOI

Bhaswati Singha Deo,

Sidharthenee Nayak, Mayukha Pal

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 100, P. 106917 - 106917

Published: Sept. 25, 2024

Language: Английский

Citations

0
Early cancer detection using deep learning and medical imaging: A survey DOI Creative Commons
Istiak Ahmad, Fahad Alqurashi

Critical Reviews in Oncology/Hematology, Journal Year: 2024, Volume and Issue: 204, P. 104528 - 104528

Published: Oct. 15, 2024

Cancer, characterized by the uncontrolled division of abnormal cells that harm body tissues, necessitates early detection for effective treatment. Medical imaging is crucial identifying various cancers, yet its manual interpretation radiologists often subjective, labour-intensive, and time-consuming. Consequently, there a critical need an automated decision-making process to enhance cancer diagnosis. Previously, lot work was done on surveys different methods, most them were focused specific cancers limited techniques. This study presents comprehensive survey methods. It entails review 99 research articles collected from Web Science, IEEE, Scopus databases, published between 2020 2024. The scope encompasses 12 types cancer, including breast, cervical, ovarian, prostate, esophageal, liver, pancreatic, colon, lung, oral, brain, skin cancers. discusses techniques, medical data, image preprocessing, segmentation, feature extraction, deep learning transfer evaluation metrics. Eventually, we summarised datasets techniques with challenges limitations. Finally, provide future directions enhancing

Language: Английский

Citations

0
An ensemble convolutional neural network model for brain stroke prediction using brain computed tomography images DOI Creative Commons
Jannatul Ferdous, Rifat Shahriyar

Healthcare Analytics, Journal Year: 2024, Volume and Issue: unknown, P. 100368 - 100368

Published: Oct. 1, 2024

Language: Английский

Citations

0