.png){kind=icon}
Return
Automatic multiple disease prediction approach using optimized convolutional neural network layers
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-026-21925-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the rapid development of computer vision and deep learning (DL), automatic and early disease prediction from medical images has drawn significant attention. Although computed tomography (CT) and chest X-ray imaging are commonly employed for lung disease diagnosis, contemporary DL-based computer-aided diagnosis (CAD) systems predominantly use computationally intensive pre-trained models, which are often constrained by redundant feature learning, limited generalization, and increased computational overhead. To overcome these constraints, this research proposes a new Automatic Disease Detection Approach (ADDA) based on an Optimized Convolutional Neural Network (OCNN) for multi-modal lung disease categorization, combining CT and chest X-ray data. Different from CNN-based frameworks, the proposed ADDA is a compact architecture jointly optimizing convolutional layer configuration, attention-guided feature refinement, sparse feature learning by L1 regularization, and sequential feature classification in a unified end-to-end framework. The medical images captured are pre-processed in a comprehensive step aiming at improving the diagnostic quality through noise reduction and contrast augmentation. The proposed Optimised Convolutional Neural Network (OCNN) automatically learns compact and discriminative feature representations using an optimised convolutional architecture with attention-guided feature refining. In the training process, we add the L1 regularisation as a penalty term in the loss function to suppress the redundant feature representations and increase the generalisation power of the network. Then the improved sequential feature vector is immediately given as input to the LSTM classifier for the correct classification of lung diseases. The proposed ADDA framework was exhaustively experimented on three publicly available CT and chest X-ray datasets with various categories of lung diseases. The experimental results show that the ADDA model outperforms many state-of-the-art DL models with an average accuracy improvement of 4.89% and a computational time reduction by 22.78%. These results prove that the suggested lightweight multimodal framework is a viable and computationally efficient option for automated lung disease diagnosis.
Keywords:
Automatic disease prediction
Convolutional neural network
Deep learning
Feature learning
Lung disease classification
Journal
IF:
3
Papers:
2.0W
Citations:
3.2W
Organization
View more organizations
Cited Papers
ShareSaveShareSaveEarly diagnosis of COVID-19-affected patients based on X-ray and computed tomography images using deep learning algorithm.png)
SOFT COMPUTING
OAAI
SOFT COMPUTINGIF2.5
OAAIShareSaveAn automatic approach based on CNN architecture to detect Covid-19 disease from chest X-ray imagesAPPLIED INTELLIGENCEOAAI
APPLIED INTELLIGENCEIF3.5
OAAIShareSaveLung Disease Classification Using Deep Learning and ROI-Based Chest X-Ray ImagesTechnologiesOAAI
TechnologiesIF3.6
OAAIShareSaveShareSaveDeep Learning Assisted Predict of Lung Cancer on Computed Tomography Images Using the Adaptive Hierarchical Heuristic Mathematical ModelIEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAIShareSave View more