Integrating Anisotropic Heat Flow and Transformer Encoders in Convolutional Neural Network for Skin Cancer Classification DOI Creative Commons
Sanad Aburass,

Maha Abu Rumman,

Ammar Huneiti

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Дек. 17, 2024

Abstract The early detection and classification of skin cancer are pivotal in improving patient outcomes reducing healthcare burdens. However, traditional deep learning models dermatological diagnostics often struggle with the nuanced differentiation lesions. This paper introduces a novel approach, integrating an Advanced Heat Flow Layer into architectures for classification, this method is centered on principles anisotropic diffusion, distinguishing itself from conventional image processing techniques by selectively smoothing areas while preserving critical edge details, essential accurate lesion identification. In our research, we utilized Ham10000 dataset, enriched data augmentation to simulate real-world variability, conducted comprehensive comparison model, featuring Layer, against several benchmark models, including Sobel Edge Detection Layer. Our integrated various layers DenseNet121, consistently outperformed these benchmarks across key metrics such as accuracy, precision, recall, F1 score, AUC, particularly augmented data, indicates significant enhancement model's ability generalize maintain diagnostic features under diverse conditions. code available at, https://github.com/sanadv/SkinCancerClassificationModels/blob/main/Models.ipynb

Язык: Английский

A hybrid machine learning model for classifying gene mutations in cancer using LSTM, BiLSTM, CNN, GRU, and GloVe DOI Creative Commons
Sanad Aburass, Osama Dorgham, Jamil Al Shaqsi

и другие.

Systems and Soft Computing, Год журнала: 2024, Номер 6, С. 200110 - 200110

Опубликована: Июнь 25, 2024

Язык: Английский

Процитировано

22
An Ensemble Approach to Question Classification: Integrating Electra Transformer, GloVe, and LSTM DOI Open Access
Sanad Aburass, Osama Dorgham,

Maha Abu Rumman

и другие.

International Journal of Advanced Computer Science and Applications, Год журнала: 2024, Номер 15(1)

Опубликована: Янв. 1, 2024

Natural Language Processing (NLP) has emerged as a critical technology for understanding and generating human language, with applications including machine translation, sentiment analysis, and, most importantly, question classification. As subfield of NLP, classification focuses on determining the type information being sought, which is an important step downstream such answering systems. This study introduces innovative ensemble approach to that combines strengths Electra, GloVe, LSTM models. After tried thoroughly well-known TREC dataset, model shows combining these different technologies can produce better outcomes. For complex Electra uses transformers; GloVe global vector representations word-level meaning; models long-term relationships through sequence learning. Our strong effective way solve hard problem by mixing parts in smart way. The method works because it got 80% accuracy score test dataset when was compared like BERT, RoBERTa, DistilBERT.

Язык: Английский

Процитировано

5
Framework for Enhanced Digital Image Transmission Security: Integrating Hu Moments, Digital Watermarking, and Cryptographic Hashing for Integrity Verification DOI
Osama Dorgham, Sanad Aburass, Ghassan F. Issa

и другие.

Опубликована: Фев. 26, 2024

Image stability is very important in a time when digital image communication essential to many fields. Modern online dangers are often too complicated for old security methods keep up with. To solve these problems, this study presents new system that combines Hu moments, watermarking, and cryptography hashing. moments create unique graphic stamp can be used check the after it has been received. Digital watermarking increases integrity of information because involves code cannot seen but detected make changes impossible. The fingerprint created with such hashing algorithms as SHA-2s other cryptographic hash functions before message transmission. This utilized verify arrives at its destination. Our combined method provides comprehensive defense against hacking, guaranteeing accuracy images sent over networks might not fully safe, structure made invisible, protecting quality while offering strong changes. We will explain how whole was put together, used, should evaluated. also discuss could situations where important.

Язык: Английский

Процитировано

1
Cubixel: a novel paradigm in image processing using three-dimensional pixel representation DOI
Sanad Aburass

Multimedia Tools and Applications, Год журнала: 2024, Номер unknown

Опубликована: Сен. 9, 2024

Процитировано

1
Cubixel: A Novel Paradigm in Image Processing Using Three-Dimensional Pixel Representation DOI Creative Commons
Sanad Aburass

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Фев. 6, 2024

Abstract This paper introduces the innovative concept of Cubixel—a three-dimensional representation traditional pixel—alongside derived metric, Volume Void (VoV), which measures spatial disparities within images. By converting pixels into Cubixels, we can analyze image's 3D properties, thereby enriching image processing and computer vision tasks. Utilizing we've developed algorithms for advanced segmentation, edge detection, texture analysis, feature extraction, yielding a deeper comprehension content. Our experimental results on benchmark datasets showcase superiority our methods in performance execution speed compared to conventional techniques. Further, discuss future applications Cubixels VoV various domains, particularly medical imaging, where they have potential significantly enhance diagnostic processes. interpreting images as complex 'urban landscapes', envision new frontier deep learning models that simulate learn from diverse environmental conditions. The integration architectures promises revolutionize field, providing pathway towards more intelligent, context-aware artificial intelligence systems. With this groundbreaking work, aim inspire research will unlock full data, transforming both theoretical understanding practical applications. code is available at https://github.com/sanadv/Cubixel.

Язык: Английский

Процитировано

0
Optimizing Customer Response Prediction in Auto Insurance: A Comparative Study of Machine Learning Models DOI
Sanad Aburass, Osama Dorgham, Jamil Al Shaqsi

и другие.

Опубликована: Фев. 26, 2024

This study examines various machine learning models to predict customer responses in the auto insurance industry. We focus on metrics like accuracy, precision, recall, and F1-score, carefully selecting threshold values balance model performance with practical business applications. Our analysis reveals XGB Classifier's superiority, achieving 99% accuracy a 98% F1-score. provide comparative of models, highlighting strengths handling complex data its efficiency compared other tested Gaussian NB Logistic Regression, which showed similar accuracies but varied precision recall. underscores importance choosing right fine-tuning it for specific industry needs.

Язык: Английский

Процитировано

0
Cubixel: A Novel Paradigm in Image Processing Using Three-Dimensional Pixel Representation DOI Creative Commons
Sanad Aburass

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Авг. 19, 2024

Abstract This paper introduces the innovative concept of Cubixel—a three-dimensional representation traditional pixel—alongside derived metric, Volume Void (VoV), which measures spatial disparities within images. By converting pixels into Cubixels, we can analyze image's 3D properties, thereby enriching image processing and computer vision tasks. Utilizing we've developed algorithms for advanced segmentation, edge detection, texture analysis, feature extraction, yielding a deeper comprehension content. Our empirical experimental results on benchmark images datasets showcase applicability these concepts. Further, discuss future applications Cubixels VoV in various domains, particularly medical imaging, where they have potential to significantly enhance diagnostic processes. interpreting as complex 'urban landscapes', envision new frontier deep learning models that simulate learn from diverse environmental conditions. The integration architectures promises revolutionize field, providing pathway towards more intelligent, context-aware artificial intelligence systems. With this groundbreaking work, aim inspire research will unlock full data, transforming both theoretical understanding practical applications. code is available at https://github.com/sanadv/Cubixel.

Язык: Английский

Процитировано

0
Parameter-Selective Continual Test-Time Adaptation DOI

Jiaxu Tian,

Fan Lyu

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 315 - 331

Опубликована: Дек. 6, 2024

Язык: Английский

Процитировано

0
Integrating Anisotropic Heat Flow and Transformer Encoders in Convolutional Neural Network for Skin Cancer Classification DOI Creative Commons
Sanad Aburass,

Maha Abu Rumman,

Ammar Huneiti

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Дек. 17, 2024

Abstract The early detection and classification of skin cancer are pivotal in improving patient outcomes reducing healthcare burdens. However, traditional deep learning models dermatological diagnostics often struggle with the nuanced differentiation lesions. This paper introduces a novel approach, integrating an Advanced Heat Flow Layer into architectures for classification, this method is centered on principles anisotropic diffusion, distinguishing itself from conventional image processing techniques by selectively smoothing areas while preserving critical edge details, essential accurate lesion identification. In our research, we utilized Ham10000 dataset, enriched data augmentation to simulate real-world variability, conducted comprehensive comparison model, featuring Layer, against several benchmark models, including Sobel Edge Detection Layer. Our integrated various layers DenseNet121, consistently outperformed these benchmarks across key metrics such as accuracy, precision, recall, F1 score, AUC, particularly augmented data, indicates significant enhancement model's ability generalize maintain diagnostic features under diverse conditions. code available at, https://github.com/sanadv/SkinCancerClassificationModels/blob/main/Models.ipynb

Язык: Английский

Процитировано

0