TriMod Fusion for Multi-Modal Lung Tumor Diagnosis

Lung cancer (LC) currently has the highest mortality rate of any medical diagnosis, underscoring the urgent need for improved detection techniques. This study highlights the critical role of automated image processing methods in improving the efficiency and accuracy of LC detection and identifies future directions for the field. The authors propose a unique method called TriMod Fusion to assist in uncovering the importance of correct tumor classification for effective patient care and management. This technique enables a more reliable and accurate classification and diagnosis of LC. A thorough examination of LIDC image data shows that the new method surpasses traditional single-modality approaches, improving the classification and diagnosis of lung tumors with higher accuracy, sensitivity, and specificity. The study presents a structural framework consisting of five sequential processes to automate the LC image recognition and classification process. Bilateral unsharp masking methods, including Gaussian, median, and adaptive, are used to improve image quality after initial image acquisition. The final stage of classification involves combining multiple neural classifiers with a hierarchical random walker and a Bayesian model. It also uses powerful segmentation algorithms and extraction methods. Findings from the third technique, which utilizes a deep recurrent neural network (DRNN) for categorization, reveal that the diagnostic precision for LC ranges from 87.5% to 97.4%, demonstrating a noteworthy enhancement in diagnostic efficiency.