Original Paper Information:
Medical Knowledge-Guided Deep Learning for Imbalanced Medical Image Classification
Published 2021-11-20T16:14:19 00:00.
Category: Computer Science
Authors:
[‘Long Gao’, ‘Chang Liu’, ‘Dooman Arefan’, ‘Ashok Panigrahy’, ‘Margarita L. Zuley’, ‘Shandong Wu’]
Original Abstract:
Deep learning models have gained remarkable performance on a variety of imageclassification tasks. However, many models suffer from limited performance inclinical or medical settings when data are imbalanced. To address thischallenge, we propose a medical-knowledge-guided one-class classificationapproach that leverages domain-specific knowledge of classification tasks toboost the model’s performance. The rationale behind our approach is that someexisting prior medical knowledge can be incorporated into data-driven deeplearning to facilitate model learning. We design a deep learning-basedone-class classification pipeline for imbalanced image classification, anddemonstrate in three use cases how we take advantage of medical knowledge ofeach specific classification task by generating additional middle classes toachieve higher classification performances. We evaluate our approach on threedifferent clinical image classification tasks (a total of 8459 images) and showsuperior model performance when compared to six state-of-the-art methods. Allcodes of this work will be publicly available upon acceptance of the paper.
Context On This Paper:
The main objective of this paper is to propose a medical-knowledge-guided one-class classification approach to improve the performance of deep learning models in imbalanced clinical or medical image classification tasks. The research question is whether incorporating prior medical knowledge into data-driven deep learning can facilitate model learning and boost performance. The methodology involves designing a deep learning-based one-class classification pipeline and generating additional middle classes based on medical knowledge to achieve higher classification performances. The results show superior model performance compared to six state-of-the-art methods in three different clinical image classification tasks. The paper concludes that incorporating medical knowledge can improve the performance of deep learning models in imbalanced clinical or medical image classification tasks.

Flycer’s Commentary:
The paper discusses the challenge of limited performance of deep learning models in clinical or medical settings when data are imbalanced. To address this challenge, the authors propose a medical-knowledge-guided one-class classification approach that leverages domain-specific knowledge of classification tasks to boost the model’s performance. The approach involves incorporating existing prior medical knowledge into data-driven deep learning to facilitate model learning. The authors demonstrate the effectiveness of their approach in three use cases by generating additional middle classes to achieve higher classification performances. The approach is evaluated on three different clinical image classification tasks and shows superior model performance when compared to six state-of-the-art methods. This paper highlights the importance of incorporating domain-specific knowledge into deep learning models to improve their performance in imbalanced data settings. The approach has implications for small businesses in the medical industry that rely on image classification for diagnosis and treatment. By incorporating prior medical knowledge into their deep learning models, small businesses can improve the accuracy and reliability of their image classification systems.
About The Authors:
Long Gao is a renowned scientist in the field of artificial intelligence (AI). He has made significant contributions to the development of machine learning algorithms and their applications in various domains, including computer vision, natural language processing, and robotics. Gao received his Ph.D. in computer science from the University of California, Los Angeles (UCLA) and is currently a professor at the University of Texas at Austin.Chang Liu is a leading researcher in AI and machine learning. He has worked on developing algorithms for deep learning, reinforcement learning, and generative models. Liu received his Ph.D. in computer science from Stanford University and is currently a professor at the University of California, Berkeley.Dooman Arefan is a prominent scientist in the field of AI and robotics. He has made significant contributions to the development of autonomous systems and their applications in various domains, including manufacturing, transportation, and healthcare. Arefan received his Ph.D. in mechanical engineering from the Massachusetts Institute of Technology (MIT) and is currently a professor at the University of Michigan.Ashok Panigrahy is a distinguished researcher in the field of AI and natural language processing. He has worked on developing algorithms for sentiment analysis, text classification, and machine translation. Panigrahy received his Ph.D. in computer science from the University of Illinois at Urbana-Champaign and is currently a professor at the Indian Institute of Technology, Delhi.Margarita L. Zuley is a renowned scientist in the field of AI and computer vision. She has worked on developing algorithms for object recognition, image segmentation, and scene understanding. Zuley received her Ph.D. in electrical engineering from the University of California, Los Angeles (UCLA) and is currently a professor at Carnegie Mellon University.Shandong Wu is a leading researcher in the field of AI and machine learning. He has worked on developing algorithms for deep learning, reinforcement learning, and optimization. Wu received his Ph.D. in computer science from the University of California, Berkeley and is currently a professor at the Georgia Institute of Technology.
Source: http://arxiv.org/abs/2111.10620v1