Project

# Title Team Members TA Documents Sponsor
54 Portable Magnetic Resonance Imaging Device
Yaokun Shi
Yujiang Han
Zexuan Cheng
Amr Ghoname design_document3.pdf
final_paper1.pdf
photo1.jpeg
photo2.jpeg
presentation1.pdf
proposal1.pdf
# Portable Magnetic Resonance Imaging Device

Team Members:
- Yaokun Shi (yaokuns2)
- Zexuan Cheng (zexuanc2)
- Yujiang Han (yujiang3)

# Problem
As image and signal processing becomes more developed and prevalent in the medical industry, advanced scanning devices are needed for the diagnosis of many diseases and such equipment is essential to many healthcare centers. However, state-of-the-art MRIs are relatively expensive and extremely large in size, which makes it difficult for smaller clinics to utilize or for professionals to quickly perform a scan on their patients on a more casual occasion.


# Solution
In order to reduce the size and cost of traditional MRI devices, we propose a portable MRI device that incorporates a non-uniform, non-linear magnetic field and two sets of RF coils. The magnetic field with unique spatial encodings will be done by rotating two individual magnets and therefore informing the RF coils about the voxel locations through predefined Larmor frequencies. The two sets of RF coils are going to be surface coil arrays that act as transmitters and receivers, respectively. The data collected from these surface coil arrays will be then analyzed and processed by our data acquisition unit, which performs current amplification as well as analog-to-digital conversion so the raw image with digital signals can be sent to our image processing unit. The image processing unit will then use parallel imaging and deep learning techniques to reconstruct images, ultimately producing an intelligible image of the scanning volume.

# Solution Components
## Base Magnetic Field

The primary purpose of this subsystem is to create a non-linear and non-uniform magnetic field in order to utilize that unique magnetic field strength to create a spatial encoding for all the surface coils in the other subsystem. To achieve this, we implemented two diametrically magnetized magnets(K&J magnet: DCCDIA) that can be rotated individually using two stepper motors(Digi-key: 2183-1208-ND), and a ferrite plate(Digi-key: 399-FPL100/100/12-BH1T-ND) is placed directly on top of the two magnets to maximize the field strength. We provide another degree of freedom by adding an extra motor on the top of this subsystem so that we can also rotate the entire base magnetic field to ensure the uniqueness of our magnetic field in our imaging volume. The rotations are controlled by individual Arduino boards that are connected to each motor, and 30 degrees of difference will be applied to each rotational operation.
## RF Transmitter and Receiver Coils

The radiofrequency coil subsystem can be divided into two components, both a 10x10 surface coil array. The transmitter coil array functions as a nuclear spin excitation trigger, which sends a signal to the target imaging volume and excites the nuclear spin of the body tissues in that region. The receiver coil array, on the other hand, detects the magnetic field change in the target region and records the strength as well as time spent for the nuclear spin to return to its original state. The signals from these receiver coils will be sent to the data acquisition unit to be further processed before visualization. The current plan is to build PCB coils because the size of these coils is more controllable, and we can also design an arbitrary number of turns and layers to satisfy specific needs.

## Control Unit

The Control Unit tells the coils when to excite the nuclear spins as well as when to collect data. This operation requires a microcontroller(potentially jetson nano) with sufficient computing power to realize the code and perform other tasks such as analog-to-digital conversion. ADC(TBD) is another crucial component in the control unit, which is needed for making the signal directly available for image processing; furthermore, due to the relatively weak field strength and certain requirements of ADC, we need to utilize a preamplifier(TBD) to increase the amplitude of the received signal to an appropriate level before digitizing. The ADC as well as the preamplifier will be embedded in the PCB, which will then be connected to the microcontroller.

## Image Processing Unit

Due to the demand for high processing power, we currently plan on using cloud computing services to reconstruct the final images. Once the digital signals are received from our data acquisition unit, the data will be sent to the cloud server through a Wi-Fi module(ESP8266). We will then use the google cloud platform for our trained neural network and send the final image to the user’s phone.

# Criterion For Success

1. Magnets can be individually rotated to arbitrary angles (30 degrees apart for each iteration) and kept in a stationary pose.

2. RF receiver coils achieve an SNR of at least 5dB.

3. The control unit is able to transmit signals at the correct frequency and perform analog-to-digital conversion within 10 seconds for each iteration.

4. Ultimately, we hope to at least get intelligible image reconstructions done on this prototypical device. Undoubtedly, the image quality will be relatively low, and the current goal is to achieve around 50%-75% image similarity.

Electronic Automatic Transmission for Bicycle

Tianqi Liu, Ruijie Qi, Xingkai Zhou

Featured Project

Tianqi Liu(tliu51)

Ruijie Qi(rqi2)

Xingkai Zhou(xzhou40)

Sometimes bikers might not which gear is the optimal one to select. Bicycle changes gears by pulling or releasing a steel cable mechanically. We could potentially automate gear changing by hooking up a servo motor to the gear cable. We could calculate the optimal gear under current condition by using several sensors: two hall effect sensors, one sensing cadence from the paddle and the other one sensing the overall speed from the wheel, we could also use pressure sensors on the paddle to determine how hard the biker is paddling. With these sensors, it would be sufficient enough for use detect different terrains since the biker tend to go slower and pedal slower for uphill or go faster and pedal faster for downhill. With all these information from the sensors, we could definitely find out the optimal gear electronically. We plan to take care of the shifting of rear derailleur, if we have more time we may consider modifying the front as well.

Besides shifting automatically, we plan to add a manual mode to our project as well. With manual mode activated, the rider could override the automatic system and select the gear on its own.

We found out another group did electronic bicycle shifting in Spring 2016, but they didn't have a automatic function and didn't have the sensor set-up like ours. Commercially, both SRAM and SHIMANO have electronic shifting products, but these products integrate the servo motor inside the derailleurs, and they have a price tag over $1000. Only professionals or rich enthusiasts can have a hand on them. As our system could potentially serve as an add-on device to all bicycles with gears, it would be much cheaper.

Project Videos