An Improved Dense CNN Architecture for Deepfake Image Detection

IEEE Access, Год журнала: 2023, Номер 11, С. 22081 - 22095

Опубликована: Янв. 1, 2023

Recent advancements in computer vision processing need potent tools to create realistic deepfakes. A generative adversarial network (GAN) can fake the captured media streams, such as images, audio, and video, make them visually fit other environments. So, dissemination of streams creates havoc social communities destroy reputation a person or community. Moreover, it manipulates public sentiments opinions toward studies have suggested using convolutional neural (CNN) an effective tool detect deepfakes network. But, most techniques cannot capture inter-frame dissimilarities collected streams. Motivated by this, this paper presents novel improved deep-CNN (D-CNN) architecture for deepfake detection with reasonable accuracy high generalizability. Images from multiple sources are train model, improving overall generalizability capabilities. The images re-scaled fed D-CNN model. binary-cross entropy Adam optimizer utilized improve learning rate We considered seven different datasets reconstruction challenge 5000 10000 real images. proposed model yields 98.33% AttGAN ^a , 99.33% GDWCT xmlns:xlink="http://www.w3.org/1999/xlink">b 95.33% StyleGAN, 94.67% StyleGAN2, 99.17% StarGAN xmlns:xlink="http://www.w3.org/1999/xlink">c that indicates its viability experimental setups.

Язык: Английский

---

Applications of machine learning and deep learning in SPECT and PET imaging: General overview, challenges and future prospects

Pharmacological Research, Год журнала: 2023, Номер 197, С. 106984 - 106984

Опубликована: Ноя. 1, 2023

The integration of positron emission tomography (PET) and single-photon computed (SPECT) imaging techniques with machine learning (ML) algorithms, including deep (DL) models, is a promising approach. This enhances the precision efficiency current diagnostic treatment strategies while offering invaluable insights into disease mechanisms. In this comprehensive review, we delve transformative impact ML DL in domain. Firstly, brief analysis provided how these algorithms have evolved which are most widely applied Their different potential applications nuclear then discussed, such as optimization image adquisition or reconstruction, biomarkers identification, multimodal fusion development diagnostic, prognostic, progression evaluation systems. because they able to analyse complex patterns relationships within data, well extracting quantitative objective measures. Furthermore, discuss challenges implementation, data standardization limited sample sizes, explore clinical opportunities future horizons, augmentation explainable AI. Together, factors propelling continuous advancement more robust, transparent, reliable

Язык: Английский

---

Review on computational methods for the detection and classification of Parkinson's Disease

Computers in Biology and Medicine, Год журнала: 2025, Номер 187, С. 109767 - 109767

Опубликована: Фев. 11, 2025

Язык: Английский

---

CNN and Bidirectional GRU-Based Heartbeat Sound Classification Architecture for Elderly People

Mathematics, Год журнала: 2023, Номер 11(6), С. 1365 - 1365

Опубликована: Март 10, 2023

Cardiovascular diseases (CVDs) are a significant cause of death worldwide. CVDs can be prevented by diagnosing heartbeat sounds and other conventional techniques early to reduce the harmful effects caused CVDs. However, it is still challenging segment, extract features, predict in elderly people. The inception deep learning (DL) algorithms has helped detect various types at an stage. Motivated this, we proposed intelligent architecture categorizing into normal murmurs for We have used standard dataset with class labels, i.e., murmur. Furthermore, augmented preprocessed normalization standardization significantly computational power time. convolutional neural network bi-directional gated recurrent unit (CNN + BiGRU) attention-based classification sound achieves accuracy 90% compared baseline approaches. Hence, novel CNN BiGRU superior DL models classification.

Язык: Английский

---

Automated Parkinson's Disease Detection: A Review of Techniques, Datasets, Modalities, and Open Challenges

International Journal on Smart Sensing and Intelligent Systems, Год журнала: 2024, Номер 17(1)

Опубликована: Янв. 1, 2024

Abstract Parkinson's disease (PsD) is a prevalent neurodegenerative malady, which keeps intensifying with age. It acquired by the progressive demise of dopaminergic neurons existing in substantia nigra pars compacta region human brain. In absence single accurate test, and due to dependency on doctors, intensive research being carried out automate early detection predict severity also. this study, detailed review various artificial intelligence (AI) models applied different datasets across modalities has been presented. The emotional (EI) modality, can be used for help maintaining comfortable lifestyle, identified. EI predominant, emerging technology that detect PsD at initial stages enhance socialization patients their attendants. Challenges possibilities assist bridging differences between fast-growing technologies meant actual implementation automated model are presented research. This highlights prominence using support vector machine (SVM) classifier achieving an accuracy about 99% many such as magnetic resonance imaging (MRI), speech, electroencephalogram (EEG). A 100% achieved EEG handwriting modality convolutional neural network (CNN) optimized crow search algorithm (OCSA), respectively. Also, 95% progression Bagged Tree, (ANN), SVM. maximum attained K-nearest Neighbors (KNN) Naïve Bayes classifiers signals EI. most widely dataset identified Progression Markers Initiative (PPMI) database.

Язык: Английский

---

A new hybrid feature reduction method by using MCMSTClustering algorithm with various feature projection methods: a case study on sleep disorder diagnosis

Signal Image and Video Processing, Год журнала: 2024, Номер 18(5), С. 4589 - 4603

Опубликована: Март 30, 2024

Язык: Английский

---

Fully automatic categorical analysis of striatal subregions in dopamine transporter SPECT using a convolutional neural network

Annals of Nuclear Medicine, Год журнала: 2025, Номер unknown

Опубликована: Март 16, 2025

Язык: Английский

---

Generating F1-Score to Predict Parkinson Disease with CNN Algorithm

Lecture notes in electrical engineering, Год журнала: 2025, Номер unknown, С. 229 - 241

Опубликована: Янв. 1, 2025

Язык: Английский

---

Machine Learning in Neuroimaging and Computational Pathophysiology of Parkinson’s Disease: A comprehensive review and meta-analysis

Asian Journal of Psychiatry, Год журнала: 2025, Номер 109, С. 104537 - 104537

Опубликована: Май 20, 2025

Язык: Английский

---

Enhanced Parkinson’s Disease Diagnosis Through Convolutional Neural Network Models Applied to SPECT DaTSCAN Images

IEEE Access, Год журнала: 2023, Номер 11, С. 91157 - 91172

Опубликована: Янв. 1, 2023

Convolutional Neural Networks (CNNs) are highly regarded in Deep Learning (DL) and have shown promising results medical image analysis, making them a leading model for Computer-Aided Diagnosis (CAD) systems. Their efficacy extends to the diagnosis of neurological disorders, including Parkinson's Disease (PD), which is typically identified through Single Photon Emission Computed Tomography (SPECT) scans. However, relying solely on visual inspection SPECT images during examinations can introduce inaccuracies due subjective factors. We propose CAD system automatic PD using pre-trained CNN models, Transfer (TL) technique, Bilinear Pooling method address this issue. The study employs several architectures, specifically Efficient-Net B0, Mobile-Net V2 custom architecture. These pre-tained architectures were originally trained ImageNet adapted current task TL technique. leveraged with bilinear pooling form, resulting three (BCNN) models. models applied pre-processed data patients Healthy Controls (HC), categorized into distinct datasets. proposed evaluated total 2720 (1360 1360 HC subjects) obtained from Progression Marker Initiative (PPMI) dataset. findings show that BCNN EfficientNet-B0-MobileNet-V2 achieved highest accuracy score 99.14%, surpassing other developed outperforming existing methods. In conclusion, provides an efficient diagnostic tool assist physicians accurate diagnoses, independent

Язык: Английский

---