The interpretability of deep learning models has emerged as a compelling area in artificial intelligence research.The safety criteria for medical imaging are highly stringent,and models are required for an explanation...The interpretability of deep learning models has emerged as a compelling area in artificial intelligence research.The safety criteria for medical imaging are highly stringent,and models are required for an explanation.However,existing convolutional neural network solutions for left ventricular segmentation are viewed in terms of inputs and outputs.Thus,the interpretability of CNNs has come into the spotlight.Since medical imaging data are limited,many methods to fine-tune medical imaging models that are popular in transfer models have been built using massive public Image Net datasets by the transfer learning method.Unfortunately,this generates many unreliable parameters and makes it difficult to generate plausible explanations from these models.In this study,we trained from scratch rather than relying on transfer learning,creating a novel interpretable approach for autonomously segmenting the left ventricle with a cardiac MRI.Our enhanced GPU training system implemented interpretable global average pooling for graphics using deep learning.The deep learning tasks were simplified.Simplification included data management,neural network architecture,and training.Our system monitored and analyzed the gradient changes of different layers with dynamic visualizations in real-time and selected the optimal deployment model.Our results demonstrated that the proposed method was feasible and efficient:the Dice coefficient reached 94.48%,and the accuracy reached 99.7%.It was found that no current transfer learning models could perform comparably to the ImageNet transfer learning architectures.This model is lightweight and more convenient to deploy on mobile devices than transfer learning models.展开更多
Through embedding modified nano-silica particles on the surface of polystyrene using the method of Pickering emulsion polymerization,a kind of nano/micro oleophobic agent named OL-1 was developed.The effects of OL-1 o...Through embedding modified nano-silica particles on the surface of polystyrene using the method of Pickering emulsion polymerization,a kind of nano/micro oleophobic agent named OL-1 was developed.The effects of OL-1 on the rock surface properties and its performance in inhibiting the oil phase imbibition into the rock were explored.The performance and mechanisms of OL-1 in improving the wellbore stability of shale gas wells were evaluated and analyzed.OL-1 could absorb on the surface of the shale core to form a membrane with a micro-nano two-stage roughness,making the surface energy of the core decrease to 0.13 mN/m and the contact angle of the white oil on the core surface increase from 16.39°to 153.03°.Compared with the untreated capillary tube,when immersed into 3#white oil,the capillary tube treated by OL-1 had a reversal of capillary pressure from 273.76 Pa to-297.71 Pa,and the oil imbibition height inside the capillary tube decreased from 31 mm above the external liquid level to 33 mm below the external liquid level.The amount of oil invading into the rock core modified by OL-1 decreased by 64.29%compared with the untreated one.The shale core immersed into the oil-based drilling fluids with 1%OL-1 had a porosity reduction rate of only 4.5%.Compared with the core immersed in the drilling fluids without OL-1,the inherent force of the core treated by 1%OL-1 increased by 24.9%,demonstrating that OL-1 could effectively improve the rock mechanical stability by inhibiting oil phase imbibition.展开更多
基金The National Natural Science Foundation of China (62176048)provided funding for this research.
文摘The interpretability of deep learning models has emerged as a compelling area in artificial intelligence research.The safety criteria for medical imaging are highly stringent,and models are required for an explanation.However,existing convolutional neural network solutions for left ventricular segmentation are viewed in terms of inputs and outputs.Thus,the interpretability of CNNs has come into the spotlight.Since medical imaging data are limited,many methods to fine-tune medical imaging models that are popular in transfer models have been built using massive public Image Net datasets by the transfer learning method.Unfortunately,this generates many unreliable parameters and makes it difficult to generate plausible explanations from these models.In this study,we trained from scratch rather than relying on transfer learning,creating a novel interpretable approach for autonomously segmenting the left ventricle with a cardiac MRI.Our enhanced GPU training system implemented interpretable global average pooling for graphics using deep learning.The deep learning tasks were simplified.Simplification included data management,neural network architecture,and training.Our system monitored and analyzed the gradient changes of different layers with dynamic visualizations in real-time and selected the optimal deployment model.Our results demonstrated that the proposed method was feasible and efficient:the Dice coefficient reached 94.48%,and the accuracy reached 99.7%.It was found that no current transfer learning models could perform comparably to the ImageNet transfer learning architectures.This model is lightweight and more convenient to deploy on mobile devices than transfer learning models.
基金Supported by the CNPC Scientific Research and Technological Development Project(2021DJ3804)Scientific Research and Technological Development Project of PetroChina Company Limited(2020E-2803(JT))China CNPC Low Carbon Strategic Forward-Looking Major Science and Technology Project(2021DJ6601).
文摘Through embedding modified nano-silica particles on the surface of polystyrene using the method of Pickering emulsion polymerization,a kind of nano/micro oleophobic agent named OL-1 was developed.The effects of OL-1 on the rock surface properties and its performance in inhibiting the oil phase imbibition into the rock were explored.The performance and mechanisms of OL-1 in improving the wellbore stability of shale gas wells were evaluated and analyzed.OL-1 could absorb on the surface of the shale core to form a membrane with a micro-nano two-stage roughness,making the surface energy of the core decrease to 0.13 mN/m and the contact angle of the white oil on the core surface increase from 16.39°to 153.03°.Compared with the untreated capillary tube,when immersed into 3#white oil,the capillary tube treated by OL-1 had a reversal of capillary pressure from 273.76 Pa to-297.71 Pa,and the oil imbibition height inside the capillary tube decreased from 31 mm above the external liquid level to 33 mm below the external liquid level.The amount of oil invading into the rock core modified by OL-1 decreased by 64.29%compared with the untreated one.The shale core immersed into the oil-based drilling fluids with 1%OL-1 had a porosity reduction rate of only 4.5%.Compared with the core immersed in the drilling fluids without OL-1,the inherent force of the core treated by 1%OL-1 increased by 24.9%,demonstrating that OL-1 could effectively improve the rock mechanical stability by inhibiting oil phase imbibition.
