Search

Manual identification of brain tumors in Magnetic Resonance (MR) images is laborious, time-consuming, and human error-prone. Automatic segmentation of brain tumors from MR images aims to bridge the gap. U-Net, a deep learning model, has delivered promising results in generating brain tumor segments. However, the model tends to over-segment the tumor volume than required. It will have a significant impact on deploying the model for practical use. In this work, the baseline U-Net model has been studied with the addition of residual, multi-resolution, dual attention, and deep supervision blocks. The goal of residual blocks is to efficiently extract features to reduce the semantic gap between low-level features from the decoder and high-level features from skip connections. The multiple resolution blocks have been added to extract features and analyze tumors of varying scales. The dual attention mechanism has been incorporated to highlight tumor representations and reduce over-segmentation. Finally, the deep supervision blocks have been added to utilize features from various decoder layers to obtain the target segmentation. The design of the proposed model has been justified with several experiments and ablation studies. The proposed model has been trained and evaluated on the BraTS2020 training and validation datasets. On the validation data, the proposed model has achieved a dice score of 0.60, 0.75, 0.62 for enhancing tumor (ET), whole tumor (WT), and tumor core (TC), respectively, and a Hausdorff 95 score of 46.84, 11.05, and 22.5, respectively. Compared to the baseline U-Net, the proposed model has outperformed WT and TC volumes in the Hausdorff 95 distance metric except for the ET volume. [ABSTRACT FROM AUTHOR]

Abstract: Financial inclusion is the delivery of financial services at affordable cost to sections of disadvantaged and low income segments of society. It has several leakages such as security threats and transaction frauds in the available system. The use of biometrics in the financial inclusion is a near perfect solution to such security threats. Such a biometric system should be fast enough to recognise/identify a subject from a large database. This paper proposes an indexing technique for a biometric database consisting of variable number of features with high dimensions. The technique makes use of geometric properties of principal components of features. Unlike existing indexing techniques, it inserts fewer features into a hash table. It reduces both computational and memory costs significantly. The geometric properties of principal components of features are found to be robust to handle translation and rotation effects. The proposed technique has been tested on palmprint biometric databases viz. PolyU [25], CASIA [26] and IITK which have 7752, 5239 and 549 images respectively. It is found to be more efficient than any other existing indexing techniques. [Copyright &y& Elsevier]

Abstract: This paper proposes a hierarchical approach to retrieve an iris image efficiently from for a large iris database. This approach is a combination of both iris color and texture. Iris color is used for indexing and texture is used for retrieval of iris images from the indexed iris database. An index which is determined from the iris color is used to filter out the images that are not similar to the query image in color. Further, iris texture features of those filtered images, are used to determine the images which are similar to the query image. The iris color information helps to design an efficient indexing scheme based on color indices. The color indices are computed by averaging the intensity values of all red and blue color pixels. Kd-tree is used for real-time indexing based on color indices. The iris texture features are obtained through Speeded Up Robust Features (SURF) algorithm. These features are used to get the query’s corresponding image at the top best match. The performance of the proposed indexing scheme is compared with two well known iris indexing schemes () on UPOL () and UBIRIS () iris databases. It is observed that combination of iris color and texture improves the performance of iris recognition system. [Copyright &y& Elsevier]