The development of an automated and accurate disease identification model to examine the stage, type, and severity of lung malignancy from computed tomography (CT) images is the more perplexing and onerous task, even for skilled radiologists, owing to the various shapes, texture, and structures of the lung nodules. Though several machine and deep learning techniques are employed to improve the effectiveness of lung nodule identification models, the predictive performance of these models still necessitates substantial enhancement to meet the real-time demands of healthcare applications. This study proposes a novel column-wise optimized extreme learner (COEL) to differentiate the CT images efficiently. The proposed COEL exploits the concept of column-wise parameter tuning whose key idea is that the tuning of the multifaceted resultant matrix is divided into fine-tuning of the matrix with a single column vector. The difficult orthogonal classification task is converted into simple least squares regression with orthogonal restraints, which can extract useful knowledge from the attribute vector of the extreme learner to create an output vector, these make COEL more regression exploration and classification aptitude. Besides, this study proposes a weighted iteratively mean-separated equalizer to enhance the quality of the scan and a boosted superpixel grouping algorithm to segment the affected regions from the CT scan and COEL to classify the sample into benign (normal) and malicious (cancerous). The performance of the intended system is cautiously analyzed on real-life databases such as LUNA16 (Lung nodule analysis 2016) dataset. The experimental outcomes gained from the COEL are compared with other existing classification models with respect to evaluation measures. The results of a comprehensive empirical analysis of the LUNA16 database divulge that the intended COEL considerably outdoes other classifiers with improved accuracy, sensitivity, specificity, precision, and Jaccard similarity score (JSS) of 98%, 97.2%, 90.8%, 96.9%, and 96.5%, correspondingly. From comprehensive experimental results, we can conclude that the proposed COEL is the better diagnostic system as compared with advanced lung cancer classification models