.png){kind=icon}
.png){kind=icon}
Return
A deep dictionary clustering approach for unsupervised image retrieval using convolutional sparse coding
.png){kind=icon}
OA
AI
DOI：10.1038/s41598-025-32982-z.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Medical image repositories have been rapidly growing due to the widespread use of imaging techniques, making manual annotation unfeasible. Efficient image retrieval systems are crucial for diagnosing diseases, planning treatments, and conducting medical research. This paper presents Deep Dictionary Clustering for Image Retrieval (DDicCIR), a novel framework that integrates deep learning with dictionary clustering for unsupervised medical image retrieval. The method employs DenseNet121 to extract image features, followed by a two-level dictionary learning process. In the first dictionary layer, the sparse representations are learned to group similar images, while the second layer refines these representations to capture higher-level abstractions and improve feature discrimination. An iterative clustering mechanism, based on k-means, updates the clusters until convergence, enhancing sparsity, reducing noise, and strengthening category separation. Experimental results on the NIH Chest X-ray and IRMA datasets show that DDicCIR achieves significant improvements in precision, recall, and mean average precision (mAP), demonstrating its effectiveness for medical image retrieval. The code is available on https://github.com/sucharithasu/DDicCIR.
Keywords:
Deep learning
Dictionary learning
K-means clustering
K-SVD
Medical image retrieval
Deep clustering
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
3.9
Papers:
27.1W
Citations:
83.5W
View more organizations