Challenges of Deep Learning in Medical Image Analysis—Improving Explainability and Trust

IEEE Transactions on Technology and Society, Journal Year: 2023, Volume and Issue: 4(1), P. 68 - 75

Published: Jan. 4, 2023

Deep learning has revolutionized the detection of diseases and is helping healthcare sector break barriers in terms accuracy robustness to achieve efficient robust computer-aided diagnostic systems. The application deep techniques empowers automated AI-based utilities requiring minimal human supervision perform any task related medical diagnosis fractures, tumors, internal hemorrhage; preoperative planning; intra-operative guidance, etc. However, faces some major threats flourishing domain. This paper traverses challenges that community researchers engineers faces, particularly image diagnosis, like unavailability balanced annotated data, adversarial attacks faced by neural networks architectures due noisy a lack trustability among users patients, ethical privacy issues data. study explores possibilities AI autonomy overcoming concerns about trust society autonomous intelligent

Language: Английский

---

A Comprehensive Review on Breast Cancer Detection, Classification and Segmentation Using Deep Learning

Archives of Computational Methods in Engineering, Journal Year: 2023, Volume and Issue: 30(8), P. 5023 - 5052

Published: July 7, 2023

Language: Английский

---

X-ray image based COVID-19 detection using evolutionary deep learning approach

Expert Systems with Applications, Journal Year: 2022, Volume and Issue: 201, P. 116942 - 116942

Published: March 30, 2022

Language: Английский

---

A hierarchical fused fuzzy deep neural network with heterogeneous network embedding for recommendation

Information Sciences, Journal Year: 2022, Volume and Issue: 620, P. 105 - 124

Published: Nov. 23, 2022

Language: Английский

---

Knowledge graph-based recommendation system enhanced by neural collaborative filtering and knowledge graph embedding

Ain Shams Engineering Journal, Journal Year: 2023, Volume and Issue: 15(1), P. 102263 - 102263

Published: April 21, 2023

Recommendation systems are an important and undeniable part of modern applications. Recommending items users to the that likely buy or interact with them is a solution for AI-based In this article, novel architecture used utilization pre-trained knowledge graph embeddings different approaches. The proposed consists several stages have various advantages. first step method, from data created, since multi-hop neighbors in address ambiguity redundancy problems. Then representation learning techniques learn low-dimensional vector representations components. following neural collaborative filtering framework which benefits no extra weights on layers. It only dependent matrix operations. Learning over these operations uses embeddings, fine-tune them. Evaluation metrics show method superior other state-of-the-art According experimental results, criteria recall, precision, F1-score been improved, average by 3.87%, 2.42%, 6.05%, respectively.

Language: Английский

---

RDERL: Reliable deep ensemble reinforcement learning-based recommender system

Knowledge-Based Systems, Journal Year: 2023, Volume and Issue: 263, P. 110289 - 110289

Published: Jan. 11, 2023

Language: Английский

---

Denoising and segmentation in medical image analysis: A comprehensive review on machine learning and deep learning approaches

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: May 17, 2024

Language: Английский

---

An interactive food recommendation system using reinforcement learning

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 254, P. 124313 - 124313

Published: June 5, 2024

Language: Английский

---

Social Recommender System Based on CNN Incorporating Tagging and Contextual Features

Journal of Cases on Information Technology, Journal Year: 2024, Volume and Issue: 26(1), P. 1 - 20

Published: Jan. 7, 2024

The Internet's rapid growth has led to information overload, necessitating recommender systems for personalized suggestions. While content-based and collaborative filtering have been successful, data sparsity remains a challenge. To address this, this article presents novel social system based on convolutional neural networks (SRSCNN). This approach integrates deep learning contextual overcome sparsity. SRSCNN model incorporates user item factors obtained from network architecture, utilizing features titles tags through CNN. authors conducted extensive experiments with the MovieLens 10M dataset, demonstrating that outperforms state-of-the-art baselines. improvement is evident in both rating prediction ranking accuracy across recommendation lists of varying lengths.

Language: Английский

---

An autoencoder-based deep learning model for solving the sparsity issues of Multi-Criteria Recommender System

Procedia Computer Science, Journal Year: 2024, Volume and Issue: 235, P. 414 - 425

Published: Jan. 1, 2024

In recent times, recommender systems have acquired significant popularity as a solution to the issue of information overload. These offer personalised recommendations users. The superiority multi-criteria over their single-criterion counterparts has been demonstrated, former are able provide more precise by taking into account multiple dimensions user preferences when rating items. prevalent recommendation technique, matrix factorization collaborative filtering, is hindered data sparsity problem user-item matrix. On other hand, it noteworthy that deep learning techniques demonstrated potential in various research domains, including but not limited image processing, pattern recognition, computer vision, and natural language processing. there surge utilisation systems, yielding promising outcomes. This study presents novel approach through algorithms mitigate issue. Specifically, autoencoders utilised uncover complex, non-linear, latent relationships between users' followed ultimately leading recommendations. proposed model evaluated conducting experiments on dataset Yahoo! Movies. According outcomes, outperforms state art methods generating accurate personalized Also, reduces up 11% from dataset.

Language: Английский

---