DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules

Lung cancer remains a leading cause of cancer-related mortality worldwide, and accurate early identification of malignant pulmonary nodules is critical for improving patient outcomes. Although artificial intelligence (AI) technology has shown promise in pulmonary nodule benign-malignant classificati...

Full description

Saved in:
Bibliographic Details
Main Authors: Shuling Wang, Suixue Wang, Rongdao Sun
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-07-01
Series:Frontiers in Medicine
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fmed.2025.1636008/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849317078531571712
author Shuling Wang
Suixue Wang
Rongdao Sun
author_facet Shuling Wang
Suixue Wang
Rongdao Sun
author_sort Shuling Wang
collection DOAJ
description Lung cancer remains a leading cause of cancer-related mortality worldwide, and accurate early identification of malignant pulmonary nodules is critical for improving patient outcomes. Although artificial intelligence (AI) technology has shown promise in pulmonary nodule benign-malignant classification, existing methods struggle with modality heterogeneity and limited exploitation of complementary information across modalities. To address the above issues, we propose a novel multimodal framework, the Dual Cross-Attention Integration framework (DCAI), for benign-malignant classification of pulmonary nodules. Specifically, we first convert 3D nodules into multiple 2D images and obtain nodule features annotated by clinical experts. These features are encoded using Transformer models, and then a dual cross-attention module is proposed to dynamically align and interact with the complementary information between the different modalities. The fused representations from both modalities are then concatenated for benign-malignant prediction. We evaluate our proposed method on the LIDC-IDRI dataset, and experimental results demonstrate that DCAI outperforms several existing multimodal methods, highlighting the effectiveness of our approach in improving the accuracy of pulmonary nodule benign-malignant classification.
format Article
id doaj-art-ae07ac0cfc634f7fa68d1b6e35e236b2
institution Kabale University
issn 2296-858X
language English
publishDate 2025-07-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Medicine
spelling doaj-art-ae07ac0cfc634f7fa68d1b6e35e236b22025-08-20T03:51:19ZengFrontiers Media S.A.Frontiers in Medicine2296-858X2025-07-011210.3389/fmed.2025.16360081636008DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodulesShuling Wang0Suixue Wang1Rongdao Sun2Department of Neurology, Haikou Affiliated Hospital of Central South University Xiangya School of Medicine, Haikou, ChinaSchool of Computer Science and Technology, Hainan University, Haikou, ChinaDepartment of Neurology, Haikou Affiliated Hospital of Central South University Xiangya School of Medicine, Haikou, ChinaLung cancer remains a leading cause of cancer-related mortality worldwide, and accurate early identification of malignant pulmonary nodules is critical for improving patient outcomes. Although artificial intelligence (AI) technology has shown promise in pulmonary nodule benign-malignant classification, existing methods struggle with modality heterogeneity and limited exploitation of complementary information across modalities. To address the above issues, we propose a novel multimodal framework, the Dual Cross-Attention Integration framework (DCAI), for benign-malignant classification of pulmonary nodules. Specifically, we first convert 3D nodules into multiple 2D images and obtain nodule features annotated by clinical experts. These features are encoded using Transformer models, and then a dual cross-attention module is proposed to dynamically align and interact with the complementary information between the different modalities. The fused representations from both modalities are then concatenated for benign-malignant prediction. We evaluate our proposed method on the LIDC-IDRI dataset, and experimental results demonstrate that DCAI outperforms several existing multimodal methods, highlighting the effectiveness of our approach in improving the accuracy of pulmonary nodule benign-malignant classification.https://www.frontiersin.org/articles/10.3389/fmed.2025.1636008/fullpulmonary nodulebenign-malignant classificationartificial intelligencemultimodalcross-attentiontransformer
spellingShingle Shuling Wang
Suixue Wang
Rongdao Sun
DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
Frontiers in Medicine
pulmonary nodule
benign-malignant classification
artificial intelligence
multimodal
cross-attention
transformer
title DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
title_full DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
title_fullStr DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
title_full_unstemmed DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
title_short DCAI: a dual cross-attention integration framework for benign-malignant classification of pulmonary nodules
title_sort dcai a dual cross attention integration framework for benign malignant classification of pulmonary nodules
topic pulmonary nodule
benign-malignant classification
artificial intelligence
multimodal
cross-attention
transformer
url https://www.frontiersin.org/articles/10.3389/fmed.2025.1636008/full
work_keys_str_mv AT shulingwang dcaiadualcrossattentionintegrationframeworkforbenignmalignantclassificationofpulmonarynodules
AT suixuewang dcaiadualcrossattentionintegrationframeworkforbenignmalignantclassificationofpulmonarynodules
AT rongdaosun dcaiadualcrossattentionintegrationframeworkforbenignmalignantclassificationofpulmonarynodules