AI 仿生眼突破有望让盲人重见光明

来自研究机构
AI 驱动的仿生眼突破有望帮助盲人重见光明
科学家利用 AI 让仿生视觉变得更智能、更精确,使为盲人恢复视力的可能性距离现实更近一步。
日期:2026年11月12日 来源:加利福尼亚大学圣塔芭芭拉分校 摘要:科学家开发出一种 AI 驱动的方法,使实验性仿生视觉技术更加精确,有助于预测一位植入脑部装置的盲人所感知到的光点闪烁与形状。通过习得个体大脑对电刺激的响应方式,这项技术有望让研究人员更接近为盲人恢复一定程度的视力。分享:
加州大学圣塔芭芭拉分校(UCSB)的 Michael Beyeler 正利用 AI 改进视觉皮层假体,也就是“仿生眼”。图片来源:Matt Perko,加州大学圣塔芭芭拉分校
人工智能有望帮助克服为盲人恢复视力过程中最大的挑战之一:让脑部植入装置产生可靠的视觉体验。来自包括加州大学圣塔芭芭拉分校在内的三家机构的研究人员已证明,AI 可以提升实验性视觉假体刺激大脑的精确度,并预测人的感知内容。
该团队利用深度学习系统,为临时植入视觉皮层的电极设计电刺激模式;视觉皮层位于大脑后部,负责处理视觉信息。通过分析大脑的响应,这项技术让研究人员对神经活动有了更强的控制力,也帮助他们更好地预判植入装置所产生的视觉感受。
这项概念验证研究在线发表于《Neuron》期刊,标志着人类朝着更先进的视觉假体——有时被称为仿生眼——又迈进了一步。这些设备最终有望通过学习每个人大脑对刺激的响应方式,而非依赖预设参数,变得更加有效。
“在抽象层面构建一个模型是一回事,”加州大学圣塔芭芭拉分校计算机科学及心理与脑科学副教授 Michael Beyeler 说。“而看着它塑造一项以真人为对象的实验,则完全是另一回事。这种从理论走向有朝一日可能真正帮助他人的能力,正是我们实验室许多工作的动力来源。”
仿生视觉如何绕过受损的眼睛
几十年来,科学家一直致力于开发能让盲人恢复部分视力的技术,但成效总体上较为有限。一些实验性设备刺激视网膜,即眼球后部感知光的组织层;另一些则尝试激活视觉系统中更靠后的部分,也就是将视觉信息传递至大脑的通路上更远处的环节。
视觉皮层假体的工作方式则不同。它们不依赖眼睛或视神经,而是将电信号直接传递到视觉皮层。对于视觉通路已受损、但视觉皮层仍能对刺激作出反应的人来说,这种方法可能提供一种潜在的解决方案。可能适用的人群包括因中风、神经退行性疾病或创伤性脑损伤而失明的人。
对于还记得看见是什么感觉的人来说,哪怕只能恢复有限的视力,也可能意义非凡。
Beyeler 特别提到那些「人生中曾有一段时间能够看见,但后来因遗传性眼病或事故而失明的人」,并补充说:「对这些人来说,想要恢复部分视力的愿望可能非常强烈。」
这项新发现建立在 Beyeler 获得的 2022 年美国国立卫生研究院(NIH)院长新创新者奖(Director's New Innovator Award)所支持的早期计算研究基础之上。这笔为期五年、金额 200 万美元的资助,支持他致力于让人工视觉系统更准确、更可预测、更有效。
科学家用脑植入装置在一名盲人身上测试 AI
测试这项技术的一个重要机会出现在 2024 年,当时 Beyeler 仿生视觉实验室(Bionic Vision Lab)的研究人员前往西班牙。他们的目标是确定其计算机模型能否提升一项已进入实验研究的脑植入装置的性能。
在 IMED Elche 医院,团队与一名 27 岁男子合作,他因创伤性脑损伤而失明。他植入了一个包含 96 个电极的装置,这些电极以阵列形式排布在他的视觉皮层内。该装置原计划在六个月内取出。
当研究人员用电流激活这些电极时,受试者会体验到被称为光幻视(phosphenes)的视觉感受。它们可能表现为光点、闪光或光的形状,有时看起来像星星或烟花。
产生这些感受是可能的,但要使其一致且可预测则困难得多。相同的电刺激并不总会引发完全相同的反应,激活特定电极也不能保证受试者一定会感知到某种特定的视觉图案。
为解决这一问题,研究人员训练了一个深度神经网络——一种旨在识别复杂模式的人工智能——来预测视觉皮层对不同电刺激参数会作何反应。该模型还考虑了刺激前即刻的大脑静息活动,从而获得关于受试者当前神经状态的信息。
借助这些预测,研究人员找出了最有可能产生特定大脑活动模式的电信号组合。
AI 让仿生视觉更精准
当科学家测试 AI 选出的刺激模式时,结果令人鼓舞。与对照方法相比,新方法生成预期神经活动模式更为准确,同时所需电流更小。
团队在考察大脑活动与受试者视觉体验之间的关系时,还发现了另一件重要的事。与刺激参数本身相比,直接从他大脑中记录到的活动模式更能预测他所感知到的内容。
这一区别很重要,因为激活大脑并不等于知道一个人会看到什么。即便研究人员能够产生预期的神经反应,他们仍需理解大脑如何把这种活动转化为有意识的视觉体验。
“工程师们自然想用对待像素的方式来对待光幻视:刺激更多电极,就应该得到更完整的图像,”Beyeler 说。“但大脑并不是这样运作的。电极之间会相互影响,神经反应会波动,我们输入大脑的东西并不一定是这个人所感知到的。挑战在于更深入地了解这种转化过程。”
这些发现表明,未来的视觉假体可能需要做的不仅仅是传递电信号。它们或许还需要监测大脑如何反应,并利用这些反馈持续改进刺激方式。
让脑植入设备学会适应每个人
实验中使用的植入设备的一个关键优势,在于它能同时完成两项任务。它可以用电流刺激视觉皮层,同时记录由此产生的神经活动。
这使研究人员能够利用大脑的实际反应来训练其 AI 模型,而不必完全依赖哪些电极被激活、以及它们输出了多少电流的信息。
这名受试者也对自己所体验到的内容提供了反馈。他会报告自己是否看到了光幻视,并在某些实验中描述其外观,包括形状、大小、亮度和颜色。
将这些报告与记录到的大脑活动进行比较后发现,相较于单纯的电刺激参数,神经反应中包含了更多关于所感知视觉感觉的有用信息。
该 AI 系统还利用静息态脑活动的测量数据,来计入每次刺激前受试者神经状态的变化。这一点可能尤其有价值,因为大脑对电信号的反应会随时间变化,甚至每天都不相同。
通过让刺激适应这些不断变化的条件,未来的假体或许能提供更一致的视觉体验,而无需让用户去适应一套固定不变的信号。
“有用的视觉假体不能依赖固定的方案,”Beyeler 说。“它必须学习每个大脑如何响应,并据此调整刺激。最终,设备应该适应人,而不是让人去适应设备。”
迈向恢复视力的合作一步
这项研究汇集了来自 UC Santa Barbara、ETH Zurich 和 Miguel Hernández University(MHU)的科学家。Beyeler 称这项努力是“一次真正的合作”。
原文较长,译文为节选;译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(AI-powered bionic eye breakthrough could help the blind see again)
from research organizations
AI-powered bionic eye breakthrough could help the blind see again
Using AI to make bionic vision smarter and more precise, scientists are bringing the possibility of restoring sight to people with blindness one step closer to reality.
Date:November 12, 2026Source:University of California - Santa BarbaraSummary:Scientists have developed an AI-powered approach that makes experimental bionic vision technology more precise, helping predict the flashes and shapes of light perceived by a blind man with a brain implant. By learning how an individual's brain responds to electrical stimulation, the technology could bring researchers closer to restoring some degree of sight to people with blindness.Share:
UCSB’s Michael Beyeler is using AI to improve visual cortical prostheses, or “bionic eyes.” Credit: Matt Perko, UC Santa Barbara
Artificial intelligence could help overcome one of the biggest challenges in restoring vision to people who are blind: getting a brain implant to produce reliable visual experiences. Researchers from three institutions, including UC Santa Barbara, have demonstrated that AI can improve how precisely an experimental visual prosthesis stimulates the brain and predicts what a person perceives.
The team used a deep learning system to develop electrical stimulation patterns for electrodes temporarily implanted in the visual cortex, the region at the back of the brain responsible for processing visual information. By analyzing the brain's responses, the technology gave researchers greater control over neural activity and helped them better anticipate the visual sensations produced by the implant.
Published online in the journal Neuron, the proof-of-concept study represents progress toward more advanced visual prostheses, sometimes described as bionic eyes. These devices could eventually become more effective by learning how each person's brain responds to stimulation rather than relying on predetermined settings.
"Building a model in the abstract is one thing," said Michael Beyeler, an associate professor of computer science and psychological and brain sciences at UC Santa Barbara. "Seeing it shape an experiment with a person is something else entirely. That ability to go from theory to something that may one day help people is what drives much of the work in our lab."
How Bionic Vision Could Bypass Damaged Eyes
For decades, scientists have worked to develop technologies capable of restoring some vision to people who are blind, although the results have generally been limited. Some experimental devices stimulate the retina, the layer of tissue at the back of the eye that detects light. Others attempt to activate parts of the visual system farther along the pathway that carries visual information to the brain.
Visual cortical prostheses work differently. Instead of depending on the eyes or optic nerves, they deliver electrical signals directly to the visual cortex. This approach could offer a potential solution for people whose visual pathways have been damaged but whose visual cortex can still respond to stimulation. Possible candidates include individuals who lost their sight because of strokes, neurodegenerative diseases, or traumatic brain injuries.
The ability to regain even limited vision could be especially meaningful for people who remember what it was like to see.
Beyeler highlighted those "who have been able to see for part of their life, but then, because of an inherited eye disease or an accident, have lost their vision," adding, "For those people, the desire may be very strong to get some vision back."
The new findings build on earlier computational research supported by Beyeler's 2022 National Institutes of Health (NIH) Director's New Innovator Award. The five-year, $2 million grant has supported his efforts to make artificial vision systems more accurate, predictable, and effective.
Scientists Test AI With a Brain Implant in a Blind Man
An important opportunity to test the technology came in 2024, when researchers from Beyeler's Bionic Vision Lab traveled to Spain. Their goal was to determine whether their computer models could improve the performance of a brain implant already being used in an experimental study.
At Hospital IMED Elche, the team worked with a 27-year-old man who had become blind following a traumatic brain injury. He had received an implant containing 96 electrodes arranged in an array within his visual cortex. The device was intended to remain in place for six months before being removed.
When researchers activated the electrodes using electrical current, the participant experienced visual sensations known as phosphenes. These can appear as spots, flashes, or shapes of light, sometimes resembling stars or fireworks.
Producing these sensations is possible, but making them consistent and predictable is much more difficult. The same electrical stimulation does not always trigger an identical response, and activating particular electrodes does not guarantee that the person will perceive a specific visual pattern.
To address this problem, the researchers trained a deep neural network, a type of AI designed to recognize complex patterns, to predict how the visual cortex would respond to different electrical stimulation settings. The model also considered the brain's resting activity immediately before stimulation, giving it information about the participant's current neural state.
Using these predictions, the researchers identified combinations of electrical signals that were most likely to produce particular patterns of brain activity.
AI Makes Bionic Vision More Precise
When the scientists tested the stimulation patterns selected by AI, the results were encouraging. The new approach generated the intended patterns of neural activity more accurately than the comparison methods while requiring less electrical current.
Another important discovery emerged when the team examined the relationship between brain activity and the participant's visual experiences. The patterns recorded directly from his brain were better predictors of what he perceived than the stimulation settings themselves.
This distinction matters because activating the brain is not the same as knowing what a person will see. Even when researchers can produce a desired neural response, they still need to understand how the brain transforms that activity into a conscious visual experience.
"Engineers naturally want to treat phosphenes like pixels: stimulate more electrodes, and you should get a more complete image," Beyeler said. "But the brain does not work that way. Electrodes interact, neural responses fluctuate, and what we put into the brain is not necessarily what the person perceives. The challenge is to learn more about that transformation."
The findings suggest that future visual prostheses may need to do more than simply deliver electrical signals. They may also need to monitor how the brain reacts and use that feedback to continually improve stimulation.
Teaching Brain Implants to Adapt to Each Person
A key advantage of the implant used in the experiment was its ability to perform two tasks. It could stimulate the visual cortex with electrical current while simultaneously recording the resulting neural activity.
This allowed researchers to train their AI model using the brain's actual responses rather than relying exclusively on information about which electrodes were activated and how much current they delivered.
The participant also provided feedback about what he experienced. He reported whether he saw a phosphene and, during certain experiments, described its appearance, including its shape, size, brightness, and color.
Comparing these reports with the recorded brain activity revealed that neural responses contained more useful information about the perceived visual sensations than the electrical stimulation parameters alone.
The AI system also used measurements of resting brain activity to account for changes in the participant's neural state before each stimulation. This could be particularly valuable because the brain's response to electrical signals can vary over time, including from one day to the next.
By adapting stimulation to these changing conditions, future prostheses might provide more consistent visual experiences without requiring users to adjust to a rigid set of signals.
"A useful visual prosthesis cannot rely on a fixed recipe," Beyeler said. "It has to learn how an individual brain responds and adapt the stimulation accordingly. Ultimately, the device should adapt to the person, not the other way around."
A Collaborative Step Toward Restoring Sight
The research brought together scientists from UC Santa Barbara, ETH Zurich, and Miguel Hernández University (MHU). Beyeler described the effort as "a true collaboration."
The project was led by co-first authors Pehuén Moure of ETH Zurich, Jacob Granley of UC Santa Barbara, and Fabrizio Grani of MHU. The supervising researchers were Beyeler, Shih-Chii Liu of ETH Zurich, and Eduardo Fernández of MHU. Their work forms part of a larger feasibility trial taking place in Spain.
Although the technology remains experimental and the participant's visual experiences were limited to flashes and shapes of light, the results demonstrate how AI could improve communication between artificial vision devices and the brain.
Rather than treating every brain as if it responds the same way, future bionic vision systems could learn from each individual and adjust their signals accordingly. That personalized approach may prove essential to developing implants that can eventually restore useful vision to people who have lost their sight.
Materials provided by University of California - Santa Barbara. Note: Content may be edited for style and length.
Pehuén Moure, Jacob Granley, Fabrizio Grani, Leili Soo, Antonio Lozano, Rocío López-Peco, Adrián Villamarín-Ortiz, Cristina Soto-Sánchez, Shih-Chii Liu, Michael Beyeler, Eduardo Fernández. Deep learning-based control of electrically evoked activity in human visual cortex. Neuron, 2026; DOI: 10.1016/j.neuron.2026.07.006
University of California - Santa Barbara. "AI-powered bionic eye breakthrough could help the blind see again." ScienceDaily. ScienceDaily, 12 November 2026. .
University of California - Santa Barbara. (2026, November 12). AI-powered bionic eye breakthrough could help the blind see again. ScienceDaily. Retrieved November 13, 2026 from www.sciencedaily.com/releases/2026/11/261110042553.htm
University of California - Santa Barbara. "AI-powered bionic eye breakthrough could help the blind see again." ScienceDaily. www.sciencedaily.com/releases/2026/11/261110042553.htm (accessed November 13, 2026).
View all the latest top news in the physical sciences & technology,
or browse the topics below:
View all the latest top news in the environmental sciences,
or browse the topics below:
View all the latest top news in the social sciences & education,
or browse the topics below:
原文为英文,正文由自动抓取并保留原文。
原文发布于 ScienceDaily AI(AI-powered bionic eye breakthrough could help the blind see again),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。