Towards trustworthy rotating machinery fault diagnosis via attention uncertainty in transformer

Journal of Manufacturing Systems, Год журнала: 2023, Номер 70, С. 186 - 201

Опубликована: Июль 29, 2023

Язык: Английский

---

Computational approaches to Explainable Artificial Intelligence: Advances in theory, applications and trends

Information Fusion, Год журнала: 2023, Номер 100, С. 101945 - 101945

Опубликована: Июль 29, 2023

Deep Learning (DL), a groundbreaking branch of Machine (ML), has emerged as driving force in both theoretical and applied Artificial Intelligence (AI). DL algorithms, rooted complex non-linear artificial neural systems, excel at extracting high-level features from data. demonstrated human-level performance real-world tasks, including clinical diagnostics, unlocked solutions to previously intractable problems virtual agent design, robotics, genomics, neuroimaging, computer vision, industrial automation. In this paper, the most relevant advances last few years (AI) several applications neuroscience, robotics are presented, reviewed discussed. way, we summarize state-of-the-art AI methods, models within collection works presented 9th International Conference on Interplay between Natural Computation (IWINAC). The paper excellent examples new scientific discoveries made laboratories that have successfully transitioned real-life applications.

Язык: Английский

---

Trustworthy clinical AI solutions: A unified review of uncertainty quantification in Deep Learning models for medical image analysis

Artificial Intelligence in Medicine, Год журнала: 2024, Номер 150, С. 102830 - 102830

Опубликована: Март 4, 2024

The full acceptance of Deep Learning (DL) models in the clinical field is rather low with respect to quantity high-performing solutions reported literature. End users are particularly reluctant rely on opaque predictions DL models. Uncertainty quantification methods have been proposed literature as a potential solution, reduce black-box effect and increase interpretability acceptability result by final user. In this review, we propose an overview existing quantify uncertainty associated predictions. We focus applications medical image analysis, which present specific challenges due high dimensionality images their variable quality, well constraints real-world routine. Moreover, discuss concept structural uncertainty, corpus facilitate alignment segmentation estimates attention. then evaluation protocols validate relevance estimates. Finally, highlight open for field.

Язык: Английский

---

A review of uncertainty quantification in medical image analysis: Probabilistic and non-probabilistic methods

Medical Image Analysis, Год журнала: 2024, Номер 97, С. 103223 - 103223

Опубликована: Июнь 1, 2024

The comprehensive integration of machine learning healthcare models within clinical practice remains suboptimal, notwithstanding the proliferation high-performing solutions reported in literature. A predominant factor hindering widespread adoption pertains to an insufficiency evidence affirming reliability aforementioned models. Recently, uncertainty quantification methods have been proposed as a potential solution quantify and thus increase interpretability acceptability results. In this review, we offer overview prevailing inherent developed for various medical image tasks. Contrary earlier reviews that exclusively focused on probabilistic methods, review also explores non-probabilistic approaches, thereby furnishing more holistic survey research pertaining Analysis images with summary discussion applications corresponding evaluation protocols are presented, which focus specific challenges analysis. We highlight some future work at end. Generally, aims allow researchers from both technical backgrounds gain quick yet in-depth understanding analysis

Язык: Английский

---

Diagnosis of brain diseases in fusion of neuroimaging modalities using deep learning: A review

Information Fusion, Год журнала: 2022, Номер 93, С. 85 - 117

Опубликована: Дек. 14, 2022

Язык: Английский

---

A review of multi-omics data integration through deep learning approaches for disease diagnosis, prognosis, and treatment

Frontiers in Genetics, Год журнала: 2023, Номер 14

Опубликована: Июль 20, 2023

Accurate diagnosis is the key to providing prompt and explicit treatment disease management. The recognized biological method for molecular of infectious pathogens polymerase chain reaction (PCR). Recently, deep learning approaches are playing a vital role in accurately identifying disease-related genes diagnosis, prognosis, treatment. models reduce time cost used by wet-lab experimental procedures. Consequently, sophisticated computational have been developed facilitate detection cancer, leading cause death globally, other complex diseases. In this review, we systematically evaluate recent trends multi-omics data analysis based on techniques their application prediction. We highlight current challenges field discuss how advances methods optimization overcoming them. Ultimately, review promotes development novel deep-learning methodologies integration, which essential

Язык: Английский

---

BotanicX-AI: Identification of Tomato Leaf Diseases Using an Explanation-Driven Deep-Learning Model

Journal of Imaging, Год журнала: 2023, Номер 9(2), С. 53 - 53

Опубликована: Фев. 20, 2023

Early and accurate tomato disease detection using easily available leaf photos is essential for farmers stakeholders as it help reduce yield loss due to possible epidemics. This paper aims visually identify nine different infectious diseases (bacterial spot, early blight, Septoria late mold, two-spotted spider mite, mosaic virus, target yellow curl virus) in leaves addition healthy leaves. We implemented EfficientNetB5 with a (TLD) dataset without any segmentation, the model achieved an average training accuracy of 99.84% ± 0.10%, validation 98.28% 0.20%, test 99.07% 0.38% over 10 cross folds.The use gradient-weighted class activation mapping (GradCAM) local interpretable model-agnostic explanations are proposed provide interpretability, which predictive performance, helpful building trust, required integration into agricultural practice.

Язык: Английский

---

Optical computing metasurfaces: applications and advances

Nanophotonics, Год журнала: 2024, Номер 13(4), С. 419 - 441

Опубликована: Фев. 2, 2024

Abstract Integrated photonic devices and artificial intelligence have presented a significant opportunity for the advancement of optical computing in practical applications. Optical technology is unique system based on functions, which significantly differs from traditional electronic technology. On other hand, offers advantages such as fast speed, low energy consumption, high parallelism. Yet there are still challenges device integration portability. In burgeoning development micro–nano optics technology, especially deeply ingrained concept metasurface technique, it provides an advanced platform applications, including edge detection, image or motion recognition, logic computation, on-chip computing. With aim providing comprehensive introduction perspective we review recent research advances computing, nanostructure methods to this work, analysis metasurfaces engineering field look forward future trends

Язык: Английский

---

Enhancing Multimodal Patterns in Neuroimaging by Siamese Neural Networks with Self-Attention Mechanism

International Journal of Neural Systems, Год журнала: 2023, Номер 33(04)

Опубликована: Янв. 15, 2023

The combination of different sources information is currently one the most relevant aspects in diagnostic process several diseases. In field neurological disorders, imaging modalities providing structural and functional are frequently available. Those usually analyzed separately, although a joint features extracted from both can improve classification performance Computer-Aided Diagnosis (CAD) tools. Previous studies have computed independent models each individual modality combined them subsequent stage, which not an optimum solution. this work, we propose method based on principles siamese neural networks to fuse Magnetic Resonance Imaging (MRI) Positron Emission Tomography (PET). This framework quantifies similarities between relates with label during training process. resulting latent space at output network then entered into attention module order evaluate relevance brain region stages development Alzheimer's disease. excellent results obtained high flexibility proposed allow fusing more than two modalities, leading scalable methodology that be used wide range contexts.

Язык: Английский

---

A framework to distinguish healthy/cancer renal CT images using the fused deep features

Frontiers in Public Health, Год журнала: 2023, Номер 11

Опубликована: Янв. 30, 2023

Introduction Cancer happening rates in humankind are gradually rising due to a variety of reasons, and sensible detection management essential decrease the disease rates. The kidney is one vital organs human physiology, cancer medical emergency needs accurate diagnosis well-organized management. Methods proposed work aims develop framework classify renal computed tomography (CT) images into healthy/cancer classes using pre-trained deep-learning schemes. To improve accuracy, this suggests threshold filter-based pre-processing scheme, which helps removing artefact CT slices achieve better detection. various stages scheme involve: (i) Image collection, resizing, removal, (ii) Deep features extraction, (iii) Feature reduction fusion, (iv) Binary classification five-fold cross-validation. Results discussion This experimental investigation executed separately for: with without artefact. As result outcome study, K-Nearest Neighbor (KNN) classifier able 100% accuracy by pre-processed slices. Therefore, can be considered for purpose examining clinical grade images, as it clinically significant.

Язык: Английский

---