Conventional single photon emission computed tomography (SPECT) relies on mechanical collimation whose resolution and sensitivity are interdependent, the best performance a SPECT system can attain is only a compromise of these two equally desired properties. To simultaneously achieve high resolution and sensitivity, we propose to use sensitive detectors constructed in a m ulti-l a yer in t e r spaced mosa ic d e tector s (MATRICES) architecture to accomplish part of the collimation needed. We name this new approach self-collimation. We evaluate three self-collimating SPECT systems and report their imaging performance: 1) A simulated human brain SPECT achieves 3.88% sensitivity, it clearly resolves 0.5-mm and 1.0-mm hot-rod patterns at noise-free and realistic count-levels, respectively; 2) a simulated mouse SPECT achieves 1.25% sensitivity, it clearly resolves 50- $\mu \text{m}$ and 100- $\mu \text{m}$ hot-rod patterns at noise-free and realistic count-levels, respectively; 3) a SPECT prototype achieves 0.14% sensitivity and clearly separates 0.3-mm-diameter point sources of which the center-to-center neighbor distance is also 0.3 mm. Simulated contrast phantom studies show excellent resolution and signal-to-noise performance. The unprecedented system performance demonstrated by these 3 SPECT scanners is a clear manifestation of the superiority of the self-collimating approach over conventional mechanical collimation. It represents a potential paradigm shift in SPECT technology development.

With the rapid development of artificial intelligence and fifth-generation mobile network technologies, automatic instrument reading has become an increasingly important topic for intelligent sensors in smart cities. We propose a full pipeline to automatically read watermeters based on a single image, using deep learning methods to provide new technical support for an intelligent water meter reading. To handle the various challenging environments where watermeters reside, our pipeline disentangled the task into individual subtasks based on the structures of typical watermeters. These subtasks include component localization, orientation alignment, spatial layout guidance reading, and regression-based pointer reading. The devised algorithms for orientation alignment and spatial layout guidance are tailored to improve the robustness of our neural network. We also collect images of watermeters in real scenes and build a dataset for training and evaluation. Experimental results demonstrate the effectiveness of the proposed method even under challenging environments with varying lighting, occlusions, and different orientations. Thanks to the lightweight algorithms adopted in our pipeline, the system can be easily deployed and fully automated.

The objective of this study was to choose the crystal surface finishing for a dual-ended readout (DER) DOI detector. Through Monte Carlo simulations and experimental studies, we evaluated 4 crystal surface finishing options as combinations of crystal surface polishing (diffuse or specular) and reflector (diffuse or specular) options on a DER detector. We also tested one linear and one logarithm DOI calculation algorithm. The figures of merit used were DOI resolution, DOI positioning error, and energy resolution. Both the simulation and experimental results show that (1) choosing a diffuse type in either surface polishing or reflector would improve DOI resolution but degrade energy resolution; (2) crystal surface finishing with a diffuse polishing combined with a specular reflector appears a favorable candidate with a good balance of DOI and energy resolution; and (3) the linear and logarithm DOI calculation algorithms show overall comparable DOI error, and the linear algorithm was better for photon interactions near the ends of the crystal while the logarithm algorithm was better near the center. These results provide useful guidance in DER DOI detector design in choosing the crystal surface finishing and DOI calculation methods.

This paper develops a MFCC-based feature for detection of epilepsy, since inspired by some methods in speech signal processing, and tests the reliability of the feature through experiments. Our experimental results show that the method using MFCC-based feature and XGBoost has a high accuracy of 99.5% in epilepsy detection, reaching the level of the state-of-the-art method. This work has some inspiration for exploring better epilepsy detection methods.