IEEE Solid-States Circuits Magazine - Fall 2019 - 44

1) provide a fundamental understanding of how the brain works, how
neurodegenerative diseases affect
our abilities, and how the brain's
computing principles can be used
to build better computers
2) create next-generation medical
devices for prosthetics and therapy for neurological disorders
3) enable brain-machine and brain-
computer interfaces (BCIs).

Understanding the Brain
The brain is the most complex organ
in the human body and a very dense
structure with approximately 50,000
neurons and 500 million synapses in
one cubic millimeter of mammalian
cerebral cortex, for a total of roughly
86 billion neurons [2]. Therefore,
understanding it is not an easy task
even with the technology we have

available today. Although research and
investigational tools for neural recording and monitoring have improved,
shrunk, and been made cheaper over
the years, neuroscientists are still able
to monitor only a few hundreds of neurons simultaneously. There is a long
road ahead to get sufficient understanding of the brain's dynamics.
An action potential (AP) is a transient (~1-5 ms) during which the neuronal membrane potential changes
rapidly and significantly, reversing
its polarity [3]. APs can serve different functions. In neuronal cell bodies
they encode information in their firing
frequency and pattern, while in axons
they serve primarily to rapidly propagate signals across distance. Therefore,
APs across the neuronal membrane
create external current flows and localized electric fields in the brain.
Parkinson's Symptoms
10 Million People

Paralysis
17 Million People

Mo
tor

tal
rie
Pa be
Lo
alam
Th

x
rte
Co s
u

Dementia
50 Million People

Aud

Temporal
Lobe

erve

itory N

Vagus Nerve

Depression
300 Million People

al
su x
Vi orte
C

Visual Impairment
1.3 Billion People

pus

Hipp

ocam

Epilepsy
50 Million People

Hearing Impairment
466 Million People

FIGURE 1: Worldwide facts about neural and other diseases affecting brain functions.
EEG

Scalp

ECoG

Skull
Dura
Cortex
White Matter

APs and LFPs

FIGURE 2: Neural signals and their locations relative to the brain anatomy. Modified from [5].

FA L L 2 0 19

IEEE SOLID-STATE CIRCUITS MAGAZINE

The combined current flows from
all active cells within a volume of brain
tissue superimpose at a given location
in the extracellular medium and generate potentials in the brain that can be
measured in several ways (see Figure 2)
[4], [5]. When electric fields are recorded
within brain tissue itself with extracellularly placed electrodes, they are called
local field potentials (LFPs). If the electric
fields are recorded from the surface of
the cortex, the method is called electrocorticography (ECOG), and it is called
electroencephalography (EEG) when the
fields are measured from the scalp.
Recording both LFPs and APs
using implanted electrodes yields the
most informative signals for studying
neuronal communication and computation. To achieve a high spatial resolution, many observation points and
short distances between the recording
sites are required. As illustrated in Figure 3, neural probes provide the best
spatial and temporal resolution, but
they cover the smallest volumes of
brain tissue, making the study of complete brain regions difficult. Therefore,
there is a need to increase the density
and number of parallel recording sites
in implantable neural probes, while
limiting their invasiveness, to enable
large-scale in vivo electrophysiology.
The Statistical Neuroscience Lab at
the University of Connecticut, Storrs,
tracks what it calls "Moore's law for neuroscience," and it calculated that progress in neural recording techniques
has enabled the number of simultaneously recorded neurons to double
approximately every seven years [6].
By 2011, this number was reported to
be fewer than 300, as shown in Figure 4
[7]. This growth is actually very discouraging when extrapolated to the
full scale of the human brain: it would
take us more than 200 years to be able
to record from all of the brains' neurons. But by using the power of silicon
and system innovation, it is possible to
break this slow trend. In recent years,
several CMOS-based monolithically
integrated neural probes have been
reported to achieve record numbers of
addressable recording sites and parallel recording channels [8]-[10].

IEEE Solid-States Circuits Magazine - Fall 2019

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Fall 2019