Andrea Goldsmith
Stanford University
MBMC Workshop
USC
December 4, 2015
Can the modeling and analysis tools of SP and Communications
Be applied to the disciplines of biology, medicine, and neuroscience
To obtain fundamental results and treatments
Three applications
 Neuroscience
 Gene Expression
 Chemical Communications
The brain as a network
Joint with N. Soltani, T. Coleman, R. Ma, J. Kim, and J. Parvizi
• Communication done through action potentials (spikes)
• Observe spike trains
0 0 1 0 1 0 0 1 0 …
0 0 0 1 0 1 0 0 1 …
0 0 0 0 0 0 0 0 0 1 0
time
• Goal: Determine physical connections between neurons
•
•
Aids in fundamental understanding of how the brain works
Can be used to study learning and degeneration
Source: http://www.tiem.utk.edu/bioed/webmodules/synapse.html
Neuronal Signaling
Directed Information
necessary for synapse to exist
•
– Not sufficient  leads to false positives
Broadcast
Relay
– Can remove false positives by observing all neurons
– Like Maximum-likelihood detection
– But we can’t observe all neurons
– What about delay inforrmation
Kim et al. (2011),
Quinn et al. (2011)
Delay Profile
≜∑
,
“bulk” of DI from X to Y
appears to be generated
here
Δ≜
→
0
5
10
15
20
Pathways through the brain
DI inference
Neuron layout
B
A
E
C
B
D
A
E
C
D
→
DI inference with delay lower bound
B
→
A
E
C
D
|
,
Constrained DI inference
B
,
|
→
A
E
C
D
|
,
We’ve developed a DMI model for the leaky integrate‐and‐fire neuron
Epileptic Seizure Focal Points
 Seizure caused by an oscillating signal moving across neurons
 When enough neurons oscillate, a seizure occurs
 Treatment “cuts out” signal origin: errors have serious implications
 Directed mutual information spanning tree algorithm applied to ECoG measurements estimate the focal point of the seizure
 Application of our algorithm to existing data sets on 3 patients matched well with their medical records
ECoG
Data
Neurocortical Silencer
• Goal: design a neurocortical silencer that cancels signals leading to epileptic seizures (Parkinson’s, depression, …)
• Current state‐of‐the‐art:
– Open loop signal injection devices using simple signals with generic (non‐personalized) parameters
– Medtronic Intercept DBS: Open loop system programmed prior to treatment – Neuropace: Detects seizure activity and injects the same generic signal for all patients. • Our proposal approach
– Use a closed loop approach similar to a phase‐lock loop to zero out (silence) electrocortical signals
– Injected signal timing, shape, and parameters will be adaptively optimized to unique electrocortical signature of each patient
Electrocortical Silencer
Waveform Generator
Adaptively change output, ε’(t), based on input, ε(t)
F(ε(t))
Personalized Electric Current Stimulation
ε’(t)
Goal: Silence the epileptic firing in the cortex
ε(t) pre‐injection
Recorded Voltage
ε(t)
G(ε(t),ε’(t))
Stimulating and Recording Electrode
Current Status
• Clinical trials on human subjects to develop the adaptive and personalized system.
ε(t) post‐injection
Microarrays as “multiuser communications”
70 genes
RNA extraction
tumor tissue
labeling
hybridization
scan
Joint with N. Zuckerman, Y. Noam, and P. Lee
Gene Expression Profiling
70 genes
RNA extraction
labeling
hybridization
scan
tumor tissue

Gene expression profiling predicts clinical outcome of
breast cancer (Van’t Veer et al., Nature 2002.)
 Immune cell infiltration into tumors  good prognosis.

Gene expression measurements: a mix of many cell
types Gene Signatures
Cell‐type
Proportion
Cell Types
Looks like CDMA “despreading”
•
Many gene expression deconvolution algorithms exist
 Shen-Orr et al., Nature methods 2010  known “C” and “k”
 Lu et al. , PNAS 2003
 known “G” and “k”
 Vennet et al., Bioinformatics 2001
 known “k”

Large databases exist where these parameters are unknown


Can we apply signal processing methods to blindly separate
gene expression?
We adapt techniques from hyperspectroscopy (Piper et al,
AMOS 2004) assuming “C”, “G” and “k” unknown
Beat existing techniques, even nonblind ones Chemical Communications
 Can be developed for both macro (>cm) and micro (<mm) scale communications
 Greenfield area of research:  Need new modulation schemes, channel impairment mitigation, multiple acces, etc. Joint with N. Farsad
Applications
York and Farsad
Vodka System
Transmitter
Custom Electrical
Switch Board
Receiver
Arduino Uno
Microcontroller
MQ303 Sensor
Adafruit LCD
Shield Kit
Arduino Uno
Microcontroller
DuroBlast
Electrical Spray
MQ-3 Sensor
MR513 Sensor
Current Work
 Concentration system has limited control on the concentration at the receiver.
 Can use acid/base transmission to decrease concentration (ISI)
 Similar ideas can be applied for multilevel modulation and multiuser techniques
Summary
 Many applications of communications and signal processing in biology, medicine, and neuroscience
 This area is in its infancy
 Need collaborators in these areas to have impact
 Timeframe and funding for this interdisciplinary research is quite different from traditional areas