The IEEE Brain Initiative eNewsletter is a quarterly online publication launched in January 2017. It features practical and timely information and forward-looking commentary on neurotechnologies and neuroengineering. eNewsletter articles can describe recent breakthroughs in research, primers on methods of interests, or report recent events such as conferences or workshops. You can contact the eNewsletter editor with any questions concerning the topic or content of your article.
An affective computing aspect on similarities and differences in emotion recognition with EEG and eye movements among Chinese, German, and French people
Wei Liu, Bao-Liang Lu
Emotions, especially facial expressions, used to be thought of as universal all around the world: we would cry when we are sad, and we would smile when we are happy. However, you might have experienced that you do not laugh after hearing a foreign joke realizing that the joke has distinct cultural backgrounds. Emotions, therefore, seem to have both universal and culturally variable components. Understanding the relationship between cultures and emotions can help us know whether emotions affect physical health in the same way across various cultures and inform us about the effectiveness of mental health interventions for patients with different cultural backgrounds. In addition, from the aspect of affective computing, a deep comprehension of cultural influences on emotions can help us build emotion recognition models for generalizing to people around the world.
Ye Tian, Cunkai Zhou, Kuikui Zhang, Huiran Yang, Zhaohan Chen, Zhitao Zhou, Xiaoling Wei, Tiger H. Tao, Liuyang Sun
Implantable flexible neural probes have been demonstrated bridging the mechanical mismatch between invasive probes and brain tissues, minimizing footprint in brain, and chronic biocompatibility . However, conventional needle-shaped flexible neural probes reported before have recording sites distributed vertically along a relatively narrow shank , which limits the lateral range in which the probes may record neural signals. Although designs with more probe shanks expand the lateral detectable range, the high implantation density reflects in increased tissue damage and surgery complexity. In this work, we developed a flexible neural probe by novel Christmas-tree structure, which has branches that are foldable along the shank by temporary encapsulation before implantation and self-stretchable after the encapsulation dissolves after implantation. The probe we developed affords increased lateral sensing range without causing extra brain tissue damage.
Kuanming Yao, Guangyao Zhao, Xinge Yu
Distinguished from conventional rigid electronics, soft electronics is becoming a novel platform for next-generation biomedical instrumentations. With advanced materials, mechanics, and structural design, soft electronics could be realized in thin, light-weighted formats and thus can be worn on or implanted in human body, and may excel in great stretchability and conformal attachment with skin or tissue, which ensures continuous and precise healthcare monitoring or therapies. Our group focuses on exploring the novel soft electronics for the applications in various fields of biomedical applications, including motion and mechanical sensing, wearable energy harvesting, dynamic temperature sensing, sweat sensing, and closed-loop human-machine interface.
Adam Li, Chester Huynh, Zachary Fitzgerald, Iahn Cajigas, Damian Brusko, Jonathan Jagid, Angel Claudio, Andres Kanner, Jennifer Hopp, Stephanie Chen, Jennifer Haagensen, Emily Johnson, William Anderson, Nathan Crone, Sara Inati, Kareem Zaghloul, Juan Bulacio, Jorge Gonzalez-Martinez, and Sridevi V. Sarma
Over 3.4 million people in the US have epilepsy and 30% of these patients have drug-resistant epilepsy (DRE), where they do not respond to medication. DRE patients are burdened by epilepsy-related disabilities and frequently hospitalized constituting around $13 billion dollars annually spent for treating epilepsy patients in the USA. Successful surgical treatment necessitates complete elimination of the brain region(s) known as the seizure onset zone (SOZ). Between 30%-70% of patients continue to have seizures 6 months after treatment due to mislocalization of the SOZ. We developed neural fragility, an electroencephalogram (EEG) marker for the SOZ, and validated it in a retrospective study of 91 patients predicting surgical outcomes using neural fragility conditioned on the clinically labeled SOZ. Fragility predicted 43 out of the 47 surgical failures correctly and had an overall accuracy of 76%, compared to the clinical accuracy of 48% (successful outcomes). Neural fragility outperformed 20 other EEG features on the same set of cross-validation samples suggesting it as a potential EEG biomarker for the SOZ.
Communicated by Dr. Jun Wang
Linran Zhao, Wen Li, Maysam Ghovanloo, Yaoyao Jia
There is an increasing realization that the majority of brain functions relate to a large distributed network of neurons that are spread over different interconnected regions of the brain. Thus, neural recording and modulation of the future will require the ability to simultaneously interface with multiple neural sites distributed over a large brain area. Traditional methods for modulating neuronal function have relied on direct stimulation by tiny electrodes, which effectiveness is undermined by the limited spatial and temporal precision with which individual cells can be selectively targeted. The emergence of optogenetic stimulation provides distinct advantages over electrical stimulation, such as cell-type specificity, sub-millisecond temporal precision, and rapid reversibility. Optogenetic neuromodulation has the potential to revolutionize the study of how neurons operate as members of larger networks and may ultimately help patients suffering from neurological disorders. Hence, we aspire to design a distributed wireless neural interface framework to stimulate large-scale neuronal ensembles over large brain areas. The distributed framework includes an array of tiny, wireless, and highly efficient implants, each of which operates autonomously to stimulate neural activities.
Communicated by Dr. Yiwen Wang
Leilei Gu, Yucheng Ding, Zhiyong Fan
“To see is to believe”. High-performance imaging devices are essential in society, particularly in the current age of Artificial Intelligence (AI) +. The biological eyes have been polished by natural selection for millions of years and their function has been verified by the diverse environment. Learning from the masterpiece of nature is therefore a shortcut to improve our manmade systems. As one of the wisest creatures in nature, human eyes are advanced image sensing systems with superiorities such as high resolution, wide field-of-view (FoV), high energy efficiency, and strong accommodations. Their high performance originates from the combined effect of a vastly flexible optical system, high-density and sensitive photoreceptor arrays, and powerful neural networks from both retina and cortex. The human eyes have a spherical shape with a hemispherical retina. A hemispherical shape matches well with the Petzval surface, which is the theoretical focal plane of the spherical lens, leading to clear and sharp imaging. In regular cameras, to mitigate the mismatching in planar structure, a delicate lens array has to be inserted to gradually bend the focal plane into quasi-flat. In our cell phones, there are 10-16 lens. With a well-designed hemispherical image sensor, high-quality imaging with a simple structure can be achieved.
Communicated by Dr. Yuxiao Yang
Shixian Wen, Laurent Itti
Lifelong learning challenges
The human brain can quickly learn and adapt its behavior in a wide range of environments throughout its lifetime. In contrast, deep neural networks only learn one sophisticated but fixed mapping from inputs to outputs. In more complex and dynamic scenarios where the inputs to outputs mapping may change with different contexts, the deployment of these deep neural network systems would be constrained. One of the failed salient scenarios is lifelong learning—learning new independent tasks sequentially without forgetting previous tasks. More specifically, agents should incrementally learn and evolve based on multiple tasks from various data distributions across time while remembering previously learned knowledge. In general, current neural networks are not capable of lifelong learning and usually suffer from “catastrophic forgetting”—learning the knowledge of the new task would overwrite the fixed learned mapping of an old task. This effect typically leads to a significant decrease of the network performances on previous tasks or, in the worst case, leads to the network completely forgetting all previous tasks.