The number of images produced per day in modern hospitals followed an exponential growth during the last decade, which has the undesirable effect to overwhelm the radiologists with image interpretation tasks. Recently, the digital form of medical images along with their standardized storage gave birth to a new domain at the crossing of computer vision and medicine: image–based computer–aided diagnosis (CAD). In this thesis, methods for building a hybrid CAD system based both on the detection of abnormal regions as well as content-based image retrieval (CBIR) are proposed, evaluated and discussed with a focus on texture analysis of high–resolution computed tomography (HRCT) images of patients affected with interstitial lung diseases (ILDs). An affine–invariant set of texture features is developed based on tailored wavelet transforms. In a second step, the clinical context of medical images is introduced in the classification framework and allowed to significantly improve the detection performance. Based on the output of the detection–based CAD, a multimodal similarity measure is introduced to enable case–based retrieval, yielding a hybrid detection–CBIR–based CAD system. The experiences and outcomes of the research presented in this thesis suggest that although computer vision is still far behind the tremendous skills of human vision, the end of its early days has been reached and multidisciplinary research efforts are required to bring the full potential of emerging computer vision technologies to the clinicians.