講演情報
[C-12-41]A 3D-Stacked Artificial Retinal Chip with CNN-Based Object Detection and Icon Stimulation
◎△Yuqi Duan1, Aoi Kataura1, Akihiro Suzuki1, Ryotaro Kawasima1, Kotaro Torazawa1, Pattaramon Thianmontri1, Yukino Sakurai1, Kyojiro Kaneko1, Kodai Koyanagi1, Takafumi Fukushima1, Koji Kiyoyama2, Tetsu Tanaka1 (1. Tohoku University, 2. Nagasaki Institute of Applied Science)
キーワード：
Artificial retina、3D-stacked chip、CNN-based algorithm、Icon stimulation、Object detection
This paper presents an energy-efficient, 3D-stacked artificial retinal chip with a lightweight CNN engine to resolve the patient "cognitive overload" caused by conventional pixel-to-stimulation mapping. Fabricated in a 0.18-µm CMOS process, the prototype executes feature extraction and pattern classification directly, translating cluttered inputs into simplified icons. The chip compresses physical footprint while maintaining total power consumption safely below the clinical intraocular thermal safety threshold.