Search

Abstract: In this paper a new algorithm for detection of suspicious mass area from mammographic images is presented. It uses histogram modification enhancement technique and a segmentation method based on minimization of inertia sum. The histogram modification filter is designed so as to be able to enhance disease patterns of suspected masses by cleaning up unrelated background clutters. Segmentation is then performed on the enhanced-image to localize the suspected mass areas using minimisation of inertia sum of images intensity classes. The proposed algorithm was tested on a database of 32 mammogramms provided by Gynaeco-obstetric and pediatric hospital of Yaoundé on which masses had previously been localised by experienced radiologists. Results show that the algorithm is able to identify masses in all cases presented with a sensibility of 94% approximately. In addition, we found out that sizes and edges of masses detected are similar to those marked by radiologists. Furthermore in some cases, we could detect some hidden masses that the radiologists were not able to mark out. [Copyright &y& Elsevier]

Three-dimensional reconstruction techniques are widely used to provide a spatial description of scoliotic spines. One example is the 3-D reconstruction method developed and routinely used at Sainte-Justine Hospital. It concerns the reconstruction of six anatomical landmarks per vertebra from the frontal and lateral radiographs: endplate centers; inferior and superior tips of both pedicles. Although this “standard” method is accurate, it requires 2 h for an expert technician to identify all anatomical landmarks on radiographs.Objective. – Therefore, this paper presents a novel and fast computed acquisition method providing a reconstruction of the spine better suited to the needs of the orthopedists during scoliosis clinics.Materials and Methods. – In this approach, five to eight points are identified on the frontal and lateral radiographs to obtain the overall shape of the spine. 2-D average predictive models, are built from a database of 791 radiographs, allowing to estimate the 2-D coordinates of the six vertebral landmarks and to obtain their 3-D coordinates from the DLT algorithm.Results and conclusions. – The evaluation of this “fast method” shows that the minimum number of points to be identified on the frontal radiograph is six for right thoracic or left lumbar curves and eight for right thoracic–left lumbar curves. Seven points need to be identified on the lateral radiograph. Furthermore, average models corresponding to a specific type and severity of scoliosis allow the 2-D mean differences between “standard” and “fast” methods to be less than 3.7 mm for endplate landmarks and 5.0 mm for pedicles. No significant difference was found between clinical indices obtained from both methods, confirming that the “fast” method is well adapted for direct clinical use. [Copyright &y& Elsevier]