62 research outputs found
Erlernen des Hashing für die medizinische Bildsuche
No full textHashing is a popular approach for performing computationally efficient approximate nearest neighbor search by means of encoding data items into sequence of bits, such that the nearest neighbor search in the coding space is efficient and accurate. This thesis explores aspects of code-consistent training of hashing forests and deep learning models for end-to-end learning of hash codes and demonstrates that such hashing models can be leveraged to perform efficient and accurate large-scale content-based medical image retrieval.Hashing ist eine Methode zur laufzeit-effizienten Nearest-Neighbor- Suche, bei der man die Daten in einer Bitsequenz kodiert, so dass die Suche im Kodierungsraum effizient und präzise durchgeführt werden kann. Diese Dissertation untersucht Aspekte des code-konsistenten Trainings von Hashing Forests und Deep-Learning-Modellen für das end-to-end learning von Hashwerten und zeigt, dass diese Hashing-Modelle dazu genutzt werden können, effizientes und präzises inhaltsbasiertes Image Retrieval für medizinische Anwendungen in großem Maßstab durchzuführen
STRATEGY FOR ELECTROMYOGRAPHY BASED DIAGNOSIS OF NEUROMUSCULAR DISEASES FOR ASSISTIVE REHABILITATION
No full textAssistive Rehabilitation aims at developing procedures and therapies which reinstate lost body functions for individuals with disabilities. Researchers have monitored electrophysiological activity of muscles using biofeedback obtained from Electromyogram signals collected at appropriate innervation points. In this paper, we present a comprehensive technique for detection of neuromuscular disease in a subject and a strategy for continuous therapeutic assessment using the Rehabilitation Assessment Matrix. The decision making tool has been trained using a wide spectrum of synthetic physiological data incorporating varying degrees of myopathy and neuropathy from beginning stages to acute. The statistical, spectral and cepstral features extracted from EMG have been used to train a Cascade Correlation Neural Network Classifier for disease assessment. The diagnostic yield of the classifier is 91.2% accuracy, 85.3% specificity and 91.35% sensitivity. The strategy has also been extended to include isotonic contractions in addition to static isometric contractions. This comprehensive strategy is proposed to aid physicians plan and schedule treatment procedures to maximize the therapeutic value of the rehabilitation process
Assessment of Driver Stress from Physiological Signals collected under Real-Time Semi-Urban Driving Scenarios
No full textAn approach for real-time stress-trend detection using physiological signals in wearable computing systems for automotive drivers
No full text
Competition vs. Concatenation in Skip Connections of Fully Convolutional Networks
No full textIncreased information sharing through short and long-range skip connections
between layers in fully convolutional networks have demonstrated significant
improvement in performance for semantic segmentation. In this paper, we propose
Competitive Dense Fully Convolutional Networks (CDFNet) by introducing
competitive maxout activations in place of naive feature concatenation for
inducing competition amongst layers. Within CDFNet, we propose two
architectural contributions, namely competitive dense block (CDB) and
competitive unpooling block (CUB) to induce competition at local and global
scales for short and long-range skip connections respectively. This extension
is demonstrated to boost learning of specialized sub-networks targeted at
segmenting specific anatomies, which in turn eases the training of complex
tasks. We present the proof-of-concept on the challenging task of whole body
segmentation in the publicly available VISCERAL benchmark and demonstrate
improved performance over multiple learning and registration based
state-of-the-art methods.Comment: Paper accepted on MICCAI-MLMI 2018 worksho
Deep Learning for Volumetric Segmentation in Spatio-Temporal Data: Application to Segmentation of Prostate in DCE-MRI
No full text
- …
