Interstitial Lung Disease (ILD) is a group of a number of chronic lung disorders involving scarring of lung tissues which may be life threatening or the cause of serious degradation in the quality of life. Thus, it is of vital importance to diagnose ILD at an early stage to fight it in the best possible way. In this article, a novel and efficient deep learning based approach for the identification of different radiological tissue patterns present in High Resolution Computed Tomography (HRCT) scans of the lung, in a slice based manner, has been presented. Since a slice based approach enables one to obtain a comprehensive view of the presence of the disease rather than a patch based piecemeal strategy, the proposed approach is inherently more effective. In this approach, Split Branch Integrated Design (SB-ID) Net, a novel deep architecture has been employed. The most significant contribution of the present work is the use of parallel processing of multi-scale features and dense connectivity patterns, effectively harnessing the strengths of both approaches to enhance feature extraction and representation. The use of multiple attention modules helps to emphasize relatively important features over less important ones, which adds to the strength of the proposed architecture. The impact of various component networks has been established through ablation studies. Performance comparisons with existing state-of-the-art alternative strategies establish the superiority of the proposed architecture. Another important contribution of the present study is the complete annotation of the publicly available database that we have used for simulation and the same will be made available free of cost to interested academic researchers.