AI for Healthcare
For some, the words “Artificial Intelligence” (AI) conjures up dystopian images of Skynet from Terminator and VIKI from I, Robot scheming to take over the world and rid the planet of pesky humans. Or maybe you’re more about the red pill and fighting the robots as Neo did in The Matrix. For me, AI doesn’t mean robots rise to the top of the food chain. Instead, AI is the future of human evolution and the natural progression of coding. AI can make human lives better and help us in ways that we couldn’t dream, especially when it comes to advances in healthcare and cancer.
In 2020, I battled cancer and won, and I did it because it was caught early by a great team of doctors and nurses. In the United States, there are almost 2 million new cancer diagnoses each year, which means about 5.5% of the US population has cancer. Cancer is the number 2 leading cause of death in US with almost 610K deaths each year. As you can imagine, detecting cancer early is really important to me and finding ways that I can contribute is also important.
Working at Oracle allows me to use our technology for good. Recently, Madhu Rao, a Principal Product Manager at Oracle, built an application called AI for Healthcare in less than a month where you can upload x-ray images to detect cancer.
Image classification is supervised learning by defining target objects (typically images or other data) and training an AI model to recognize them using labeled examples. In recent years, image classification has seen a rise in use as a tool for medical diagnosis. Many physicians and researchers are using deep learning AI models to classify medical images, such as X-rays, CT Scans and MRIs, into distinct categories to augment existing diagnostic techniques. The techniques aid doctors and research activities in their decision-making. However, they don’t replace traditional methods of diagnosis and treatment by doctors. Many of these AI outcomes still need human judgement and decision-making.
Currently, the AI for Healthcare app detects breast and lung cancers, COVID-19 and Pneumonia, but the model can be trained to detect other types of cancers, such as Prostate cancer, Pancreatic cancer, Thyroid cancer, Brain tumour, Uterine cancer and many more.
Below is an x-ray image of a breast from a public data source found by doing a simple search on “cancer x-rays datasets” on Google. The app processes the image, compares against the trained model loaded into Oracle Cloud Infrastructure (OCI), and provides a percentage of probability of breast cancer. The front-end is built on APEX and the backend is powered by Oracle’s Autonomous Database. Using OCI’s Vision to detect anomalies in tissues that are indicators of cancer, it can give you a probability score of cancer or normal tissue present in the x-ray image. In the image below, the left side displays the probability score, which has a probability score of 97% of cancer and 3% normal, and the right side shows the x-ray image of the breast tissue.
Clicking on the Breast Cancer 3D Model link also shows a 3D image and details about breast cancer:
If you click the “Play” button, it will load the 3D model so you can view from any angle:
The AI for Healthcare app also includes a document reader using OCI’s Document Understanding that scans a document (the document is a CVS receipt in this example below), extracts text, loads them into your database tables, as well as other key data from document files through APIs and command line interface tools to be displayed in the app. You can also easily create an expense report from the receipt and who likes doing expenses? Not me.
In case hearing or seeing is a challenge, there is also a medical transcription feature in the AI for Healthcare app that will allow you to record speech, or upload a .wav file to convert to text.
AI also responds back to an audio message as seen in the example below:
For example, when patient says, “Doctor, I have a severe headache. Please help me. Thank you.“ AI responds back with “I’m sorry to hear about your headache. To help relieve the pain, I suggest taking an over-the-counter pain reliever such as ibuprofen or acetaminophen. You should also try to rest in a quiet, dark room and drink plenty of fluids. If the headache persists, please consult your doctor.” Nifty, right?
Patients can also chat with an AI-based doctor. For example, patients can ask, “How do I know if I have breast cancer?”. AI responds with, “It is important to see your doctor if you are concerned about breast cancer. Your doctor will be able to perform a physical exam, as well as order further tests like a mammogram, ultrasound, or biopsy to determine if you have breast cancer.”
We’ve even integrated a chatbot feature so you can ask our chatbot, affectionately known as Dr. Artie, anything! Oracle Digital Assistant powers Dr. Artie, who is constantly learning. You can even ask Dr. Artie a question using your voice if you don’t want to type.
Using Oracle Spatial, we’ve also added a map feature that displays cities with hospitals, as well as their specializations so you can find a hospital near you.
Electrocardiography is the process of producing an electrocardiogram (ECG or EKG ), a recording of the heart’s electrical activity through repeated cardiac cycles. It is an electrogram of the heart, which is a graph of voltage versus time of the heart’s electrical activity using electrodes placed on the skin. These electrodes detect the small electrical changes that are a consequence of cardiac muscle depolarization followed by repolarization during each cardiac cycle (heartbeat). Changes in the normal ECG pattern occur in numerous cardiac abnormalities, including cardiac rhythm disturbances.
AI for Healthcare also demonstrates how Electrocardiogram (ECG) Reports along with OCI Anomaly Detection can be used to Detect & Diagnose Cardiovascular disease.
In addition, you can incorporate Blockchain tables in the AI for Healthcare app. There are many scenarios where Biometric information, a single source of truth, can be stored in blockchain tables where outsiders do not tamper with biometric data.
“ By combining biometrics and the blockchain, tech leaders can create solutions to keep a user’s identity inside a secure distributed ledger system, achieving complete human control. For example, biometrics like heartbeat and voice signals could prevent people from stealing someone’s identity in the financial sector. It could be quite easy to condition entry to the blockchain to the fail/pass test of being a human without the need for revealing one’s identity in what is called a pseudonymous manner. In other words, inventors can simply create a condition test to validate users through physiology,” says Martin Zizi, Former Forbes Councils Member, Forbes Technology Council on forbes.com.
You will get an error message that the record cannot be updated in a Blockchain table since biometric data shouldn’t be tampered with and ensures adherence to security protocols.
This is also integrated with Oracle Analytics Cloud service for Clinical Data Analysis, Protein Analysis v/s Peptide Targets, and Cancer Open Research Data. Below are examples of each research data:
Below is the full architecture of the solution, which took less than a month to build:
Check out the links below to learn how you can build this in your own environment!
Workshop is available at: https://bit.ly/O_AI_healthcare
Check out the video to learn more: AI for Healthcare Demo and Architecture — YouTube
And follow me to get updates!
