The SWITCH PATHS-UP program hit the mark in terms of research, programming opportunities, assignments, tutorials, and guest speakers. Although the program was entirely virtual, it didn’t disappoint in terms of “rigor.”The basis for the SWITCH PATHS-UP program centered around Healthcare and Computer Science.
Healthcare is a topic of debate in all parts of society. One mainstay of healthcare is “affordability.” There are scores of individuals in the United States who do not have health because they can’t afford it, or if they have healthcare, it’s through third-party vendors with menial coverage, or it’s Medicaid or Medicare. Unfortunately, people who live in underserved communities have to make life and death decisions when it comes to healthcare.
Some of the projects, programs, and devices that were availed during the program centered around healthcare are in development or are in use today. It’s excellent news for people that live in underserved communities. Because access to healthcare programs is another obstacle people have to face.

We had the privilege of meeting Dr. Temiloluwa Prioleau. She is an Assistant Professor of Computer Science at Dartmouth College. Her goal is to leverage her expertise in Computer Science and Health Care to assist underserved communities with healthcare issues. Her work is outstanding!

We are living in unprecedented times. Computer software developers, doctors,  scientists are racing against the clock to create a vaccine and a cure for COVID19. One common element amongst all doctors, researchers, software developers, and scientists is “data.” We live in a “data-driven society. With the right type and amount of data, patterns could be discovered to help medical answer tough questions and problems. Through data, medical techniques and procedures could be improved. Maybe a vaccine or are a cure for COVID19 could be discovered!
Throughout the PATHS-UP research program, I have become more acutely aware of how “powerful” research and data play a significant role in solving problems.
Now that the SWITCH PATHS-UP program has ended, my computer science journey just got a bit more interesting. Before I started writing this last blog post, I enrolled in an online Python course to reinforce what I just experienced and to take me further and deeper into Computer Vision, OpenCV, AI, and Machine Learning. Keep Learning!

We wrapped up our last week with 2 amazing presenters! First was Jasper Tan and his focus was on Privacy in the Age of Data. He told us how in order to train machine learning networks and gather all this data a bg issue arises. How do you protect/prevent the sharing of private and sensitive data to undesired users?? He mentioned how even retracting someones name and address left “quasi-identifies” that can narrow you down.

Last but not least we had Temiloluwa Prioleau present to us about using data to understand, monitor and improve health outcomes.

We got a chance to revisit how machine learning can help monitor diabetes and ultimately make a more user friendly device.

We are officially at the last blog post! It feels like yesterday I got the email informing me I got invited to the SWITCH program and jumped around celebrating lol. I’m eternally grateful to RICE and SWITCH for giving me the opportunity and privilege to partake in this program. I feel like I learned stuff that made me a better programmer and teacher. I literally had to rearrange my bookmarks since I was saving absolutely every resource that got thrown my way. I think a big hurdle, at least for me personally, was the vast amount of results that pop up when you google “how to learn machine learning.” With this program we got specific tools, packages, and readings to tie everything together. One thing I was fearful of when I started teaching was that I would be teaching outdated things. Like that one professor who is still teaching Python 2 somewhere in america. This program helped alleviate that fear and if anything kind of reassure me that one never has to stop learning. and you don’t need to re-enroll into school to keep up.

In the age of information ignorance is a choice.

A Pause, A Break

While I may not be able to treat COVID, I’m happy that my family remains in good spirits. I face-timed my little sisters today and they seemed to be doing well. I loved hearing their voices earlier today. It helped me stay grounded and gave me the energy to get back to work. Week 4 is over and I can’t believe it? SWITCH is almost over. *insert cringe emoji* Also, I tried to do the snapchat thing from last week, ’twas not nice. But, I’m not giving up! My grandma says if she hasn’t given up, neither can I!

Continue reading →

Red Eye

This week started with a lot of excitement as we were able to finally get our eye-tracking program working with some assistance (Leon, Daniel & Chris- shout out!). I’ve attached my first try at this- it was remarkably hard to actually get the right threshold. Funny enough, my fellow RETs seem to have in the range 20-50, while I was actually 77. It might be interesting to explore how the threshold ties in with the lighting in the room. Below is my first attempt at eye tracking, if you look closely you’ll see a red circle on the left pupil. Our weekly article informed us that this particular tracking is classified as a geometric gaze model- due to the fact that the computer is trying to find the circular shape of the pupil.

First attempt at eye-tracking.

Attempt at Snapchat Filter

Started my feeble attempt at making a googly eye snapchat filter. I realized I needed a different version of Python to install Dlib. I’ll be continuing this project next week as I would love to learn this- my students would be thrilled to do this- even if it is complicated.

Brain·i·ac

Our mentor this summer, Mary shared her research with us. Her research focuses on a microscope called RAPID- which stands for Real-time Artificial-intelligence-enabled Pathology for Intraoperative margin Detection. This talk was quite interesting, but it had concepts that were way above my intelligence level. I’ve attached a picture to show how the RAPID microscope works.

How RAPID works

Here is how RAPID microscope does in comparison to a conventional microscope. It is visible that the quality of the slides is clearly better with a RAPID microscope.

Visibly clearer results with a RAPID microscope.

Obama TIDBIT: Mary shared an interesting article where Obama’s face was almost fitted with white features due to a dataset that only included Caucasian males- it is important for us to realize that our results will only be as efficient as our dataset, if our dataset is biased- our results will be too. Therefore, when collecting samples we have to ensure that the dataset is of a wide variety.

Brady’s research was on cancer prevention in rural areas. He was able to put his bio-engineering degree to excellent use. His project focused on a mobile truck pap smear program that was offered to assist in the early detection and treatment of cervical cancer in women living in the rural areas. Unfortunately, it is noted that 40% of women do not come to the Central hospital for diagnostic follow-up and treatment due to the distance barrier.

Here is where the Brady’s research team is able to design new technologies that are not only cost effective but reduce result time in cervical cancer detection. While the standard methods for cervical cancer prevention are pap smear test, colposcopy, histopathology, and loop electrosurgical excision procedure- the research is able to design some scaleable methods such as rapid HPV test, mobile colopscopy, mobile microendoscopy, and portable thermocoagulator which can be used by a single practitioner in a single visit in conjunction with low-cost training kit.

Brady’s main focus

Focus on Presentation

The rest of the week was focused on creating our first draft for the presentation next Friday- stay tuned!

     Week four has been a blast! I can’t believe so much time has passed. We had several guest speakers who were Rice Graduate Students. They visited our zoom meeting to chat about their work. Mary Jin and Brady Hunt are working on groundbreaking research centered around healthcare. 

     Mary Jin showcased her work on using imaging equipment to locate cancerous cells in tissue samples. Her talk was of particular interest to me not only for the “health” aspect but also for the imaging aspect. The level of detail in the images she shared was astounding. 

      Brandy Hunt’s research centered around making healthcare (screening for cervical cancer) in women accessible and affordable. Brandy understood that there are numerous barriers preventing women from getting proper screening. One is proximity to proper healthcare; the other is cost, among others.  There is a disproportionate number of people all over the World, who don’t have access to quality health. Cost is another concern that delays proper healthcare for these same individuals. He and his team worked took a mobile health unit to bring screening to distant villages in Brazil. His end goal is to create low-cost handheld devices that could be used anywhere, and at any time to test for cervical cancer. 

     The main focus this week was completing research for my presentation. I have become very interested in the applications for computer vision. One such application is “Eye Tracking and Gaze Detection.”  A question had arisen during this week that made me think a little harder. What is the difference between Gaze Detection and Eye Tracking? Gaze Detection is where a user’s eyes are focused when sitting in front of a monitor—Essentially, where on the screen users are looking. Eye-tracking is the process of measuring the motion of an eye relative to the head.

This entire time I though a was working on Gaze Detection. In reality, I was working on Eye Detection. see images below. The red-eye tracker is following my eye movement not necessarily where am I looking.

My research will be incorporated into a CS lesson I am working on for the fall semester. My students will create a program that will track their eye movement; while simultaneously using a clicker built from a Lego EV3 (student-created and programmed) to test their peripheral-vision.

Eye Tracking © C. Franklin

Eye Tracking © C. Franklin

We kicked off this week with a presentation from our fearless leader, Mary Jin who discussed her work in developing and testing the RAPID (Real-time Artificial-Intelligence-enabled Pathology for Intraoperative margin Detection) microscope for cellular imagery. The dilemma discussed was easy to understand. A surgeon is operating on Johnny Bravo’s cancerous tumor. They do the best they can to excise the tumor completely. So how do they know they got enough? Well, obviously they need to cut the tumor into cross-sections. Give the cross-sections to the friendly neighborhood pathologist that everyone plays croquet with on the weekends. Wait for the pathologist to use a handy dandy conventional microscope to examine the cellular structure of the edges of the cross-sections. Only to find out there are still abnormal cells on the perimeter of the tissue which means there is a high probability that cancerous tissue was not completely removed and Johnny Bravo has to be put back under.

The RAPID microscope aims to eliminate the need for relying on slower and costly methods for cellular imagery by developing and extended depth of field microscope that leverages a neural network and machine learning to deliver stunning results. The future of “RAPID” cellular microscopy is here!

We also were enlightened by Brady Hunt. A graduate student from Dr. Rebecca  Richards-Kortum Bioengineering lab. He shared his inspiring work in taking on the challenge of providing cancer prevention treatments for women in rural areas at higher risks of developing and dying cervical cancer. The collaborative efforts of folks from Rice, Barretos Cancer Hospital in northern Sao Paulo, Brazil, and Baylor College of Medicine are nothing short of inspirational. Teams of engineers, research scientists, and doctors from different parts of the world coming together to holistically revamp standard methods of care is how we as a society can change the future for underserved communities. By developing new methods that use cost-effective technology and designed to scale in low-resource settings, I am optimistic that Brady and Mary’s research will have a positive impact on many lives!

It feels like we just started this program and we’re already wrapping up. This week was the most challenging for me. Our reading for this week had less of a learning curve and I was able to understand most of it as I was reading (unlike the readings last week lol.) It pretty much covered Gaze360, a gaze-tracking dataset and method for 3D gaze estimation. It covered how they gathered the dataset using a 360 degree camera (with the top turned off) and took it to gather the data in outdoor locations. I was surprised at how small of a dataset it was with only 238 subjects, but also how effective it was. I was under the impression that machine learning datasets had to be in the thousands.

For our coding project we had to track our eye using python. According to Anil this is the first step in gaze detection and it makes perfect sense. Can’t see where the eyes are looking if you can’t accurately tack the eyes themselves. We followed a blog post by Stepan Filonov and while his explanations and procedure were easy to follow at first he kept changing the variable names so you couldn’t just copy and paste his code. I’m not sure if this was done on purpose or not? Also towards the end it seemed like it was more rushed and the new code wasn’t specified where it should go. Either way with some debugging and retracing I was able to begin tracking my eyes. Although I had to adjust the threshold first. According to Stepan:

The way binary thresholding works is that each pixel on a grayscale image has a value ranging from 0 to 255 that stands for its color. You pass a threshold value to the function and it makes every pixel below the value 0 and every pixel above the value the value that you pass next, we pass 255 so it’s white.

after changing the threshold to 51. I was able to tracking my eyes with acceptable results.

We also had a presentation by Mary! She told us about RAPID: Real-time Artificial-intelligence enabled pathology for intraoperative margin detection. basically shes going to save the world! The problem is that when doctors are evaluating tumors doctors have to make a cross-section of the tissue, stain them, and the analyze them. This is a process that apparently takes day. But with mary’s microscope they can scan in under around 10 minutes and accurately analyze the nuclei and look for anything that’s too big, dense, or erroneous.

Personal Plights

This week has been a struggle. I had mentioned in a prior post how crazy it felt. Well, it got crazier. Now, my whole family has had to quarantine and I’m walking around looking like a plague doctor trying to care for them before I head back to Austin. Sometimes this year has felt like a waking nightmare. The thought of losing my grandmother and my family to this virus has brought those feelings to their peak. I don’t think I’ve ever cried so much in so few days. The only consistent thing throughout all of this has been PATHS UP with RSTEM but, even then, it has felt very inconsistent this week do to the family emergencies. I hope next week is better and that my family makes a fully recovery.

DeepMind, Lord of the Rings, and Vaccines

Straight from the recesses of my decomposing mind, I bring you my latest unhinged theory based on a conversation with special Guest Speaker, Arun Ahuja, that we had courtesy of one of our amazing cohort members–thanks Daniel! Arun was there to speak to us about Google DeepMind and all of the possibilities that were related to it, the video games, the labyrinths, the rewards systems that were designed specifically to allow the system to learn on its own. It was nothing short of amazing. Then, I had a thought about one of my favorite video games that I’d played: “Middle Earth:Shadow of War”. Regardless of how familiar you might be with this installation in the Lord of the Rings series of games, the premise is simple, as the main character you seek to conquer the world by strategically assassinating, taking control of, or collaborating with characters in the game. But, they use something called the Nemesis System to make this possible–and this is how it relates back to DeepMind–in which every single possible outcome is different and the non-playable characters can make their own choices to betray, join, or resist. The neural network system that Arun described sounded like how I imagined the Nemesis system would work, accounting for every eventuality, every move, like the AI that defeated the top Go player. So, I took this comparison further, I looked for confirmation that the Nemesis system in the game was somehow like a CNN but could not find concrete evidence to support my theory. (I spent five hours looking for it while caring for my family) I continued with the course though, and in the conversation with Arun asked if there was any credence in applying this DeepMind approach, with this Nemesis System outcome that I’ve seen and applying it to virology. My logic here was: if we can let an AI teach itself how to combat a virus and generate a subsequent vaccine, then allow the AI to account for every iteration in RNA mutation at any given sequence, we should—in what I’m sure is my flawed theory—be able to generate a vaccine for such things liked COVID-19. Then, we can expand upon that concept further and generate possible vaccines for viruses that do not yet exist.

Boom and Bust: Kaggle & Tensorflow

I felt really good about the Kaggle course working with Pandas for Python. There was definitely a familiarity to it in that it felt like commands for SQL and for using Pivot Tables for data. I think it was one of the few times I didn’t feel completely lost in the journey. Granted, I feel like at this point we have been learning so much in a short amount of time that I should still be grateful for what I can understand. Tensorflow has not been nice at all though. It has probably given me more grief than PyCharm did in the beginning. First it wouldn’t download, then it was corrupted, and finally it shut off my laptop for no reason. So, we are going to re-attempt Tensorflow tomorrow and Sunday because this week really left me with no time to do much else.

… seriously. I could probably write 250 words on how unjust it is to do a single blog post covering the topics we hit on this week. It took me a day just to process and be able to sit down and write about this week. All three meetings we had this week deserve their own blog post but alas I digress… and I’m already 67 words in.

Implicit Bias Training

Monday we sat down with the wonderful Dr. Kristen Antoine-Morse to discuss principles for nurturing diverse learning environments. The two things that resonated with me the most were around the discussions we had regarding cultural identity and implicit bias. 

Principles for Diverse Learning Environments

To build a better classroom for all learners it’s imperative that we, as educators, take the time to know our students. We must strive to connect with them so that they feel valued as individuals and safe to be who they identify as. Cultural identity extends beyond race and gender and it’s too easy to look at our students and generalize things about them based on our implicit biases. 

Cultural Identity

If we do not reflect on our own biases, then we are fostering an environment that inevitably leads to discrimination. The conversations will not be easy but we must rip off the metaphorical bandaid if we ever hope to heal the gaping wounds of systemic racism and wealth inequality. Against the backdrop of the COVID-19 pandemic,  Black Lives Matter, and the immigration crisis, the start of the 2020-21 school year has the potential to be the most difficult year for educators and students across the world. The silver lining is that it can be a turning point for our country’s future. I’m determined to play my part and hold my colleagues to account.

This week I got to research Machine Learning!

I chose to follow Google’s Machine Learning Crash Course for a couple of reasons: 1) it uses Tensor Flow API without requiring me to download Tensor Flow in my machine, 2) it uses Colab notebooks , which are AWESOME and 3) it’s short but very detailed and comprehensive.

Since this was a short week, and we still consumed most of our day with meetings, I’m not even halfway through the tutorial, but I have learned so much already.

I really loved the way they introduced Machine Learning with Linear Regressions and it got me thinking, do teenagers realize they are learning a little Machine learning in 9th grade?

What if we taught students the application of what they are learning, rather than just solve for x? I realized also, that there aren’t many RET programs out there, how would a teacher know about Algebra being a cornerstone bridge to Machine Learning without a research opportunity?

Moving on.

I learned about loss, specifically L2 loss.

“Loss is the penalty for bad prediction. If a model’s prediction is perfect, loss is zero”. – Google ML Crash Course

L2 Loss Formula

This week we also got to hear from two speakers: Grad student Yongyi Zhao and  Senior Research Engineer, Arun Ahuja.

Arun talked about the work at Deep Mind and how they are applying Deep Learning in Games. All this work and research is truly fascinating and I look forward to diving more into Machine Learning Next week.

For now, we will take a little break for the 4th of July!