22 November 2023 in Journal of Geophysics and Engineering
Hydraulic fracturing, a significant contributor to seismic activity within and around operational fields, has been extensively used in shale gas production. Magnetotelluric (MT) sounding is an effective geophysical tool for identifying high-conductivity fluid-filled and/or molten regions. In this study, we deploy a dense grid of rectangular MT sites to investigate the 3D geoelectrical resistivity structure beneath the Weiyuan shale gas block (WSGB) and subsequently examine the causes of seismic activity. The resistivity data, obtained through 3D inversion accounting for topography using ModEM, reveals a shallow low-resistivity layer (<10 Ω-m) within the WSGB, ranging from ∼2 to 5 km in depth. This layer exhibits multiple isolated areas with very low resistivity (<5 Ω-m), indicative of fluid-filled zones associated with hydraulic fracturing or shale gas-bearing formations. In the northwestern WSGB, the Weiyuan anticline presents a high-resistivity dome extending possibly to depths beyond 20 km, without extending beyond the northern boundary of our study area. Conversely, the sedimentary zone in the southeastern WSGB displays a low-resistivity feature, with an extremely low-resistivity center (<1 Ω-m). Since a consistent high-resistivity zone exists beneath each fault and its top depth is <5 km, so faults might not extend downward below 5 km. Earthquakes with magnitudes (ML) of 3.0 or higher predominantly occur close to the faults, when considering industrial production data, we found a noteworthy correlation between earthquakes with ML < 3.0 and annual shale gas production within the WSGB. Tectonic faulting is not the leading cause for ML < 3.0 earthquakes but likely the primary contributor to seismic events with ML ≥ 3.0.
17 November 2022 in arXiv
Automated polyp segmentation technology plays an important role in diagnosing intestinal diseases, such as tumors and precancerous lesions. Previous works have typically trained convolution-based U-Net or Transformer-based neural network architectures with labeled data. However, the available public polyp segmentation datasets are too small to train the network sufficiently, suppressing each network's potential performance. To alleviate this issue, we propose a universal data augmentation technology to synthesize more data from the existing datasets. Specifically, we paste the polyp area into the same image's background in a spatial-exclusive manner to obtain a combinatorial number of new images. Extensive experiments on various networks and datasets show that the proposed method enhances the data efficiency and achieves consistent improvements over baselines. Finally, we hit a new state of the art in this task. We will release the code soon.