Considering Learnability: Brainome

 Was just introduced to a new company:    Brainome    An example problem addressed,  indicating breadth and possibilities :  

" ... We've been working with a genomics research group to analyze the cancer atlas which aggregates gene expression data from patients diagnosed with 33 different kinds of cancers. We combined the cancer atlas with normal (healthy) samples from the GTEx database to create a 34 class / 21,000 column / 11,000 row data set. The goal is to find genes that are consistently over-expressed in cancer patients and, so far, we've identified ~3 dozen that are good candidates for further testing. The long term goal is an early detection blood test for certain types of cancer.  ... " 

Measure and improve the learnability of your data

Do I have enough data?

Do I have the right data features?

Which of my data features are the most impactful?

Will my model overfit?

What kind of ML model will work best with my data?

What is my accuracy vs generalization curve?

Worth a look, See their Blog      And FAQ:  

What Blockchain Could Mean for Your Health Data

Thoughts about blockchain and health data

What Blockchain Could Mean for Your Health Data
by Don Tapscott and Alex Tapscott  in HBR

Big data is perhaps the most powerful asset we have in solving big problems these days. We need it to track and trace infection, manage healthcare talent and medical supply chains, and plan for our economic futures.

But how can we balance data and privacy? Legislation and regulation of big data such as the European Union’s General Data Protection Regulation (GDPR) and California’s Consumer Privacy Act are partial measures at best. Regulators and pundits have focused so much on the demand side of the data equation — that is, on the use or sale of private citizens’ data in corporate applications like Facebook, Google, and Uber without the individuals’ awareness — that they’ve failed to look at the supply side of data: where data originates, who creates it, who really owns it, and who gets to capture it in the first place.

The answer is you do. All these data are a subset of your digital identity — the “virtual you,” created by your data contrail across the Internet. That’s how most corporations and institutions view you. As Carlos Moreira, CEO of WISeKey, said, “That identity is now yours, but the data that comes from its interaction in the world is owned by someone else.”

It’s time we started taking our personal data as seriously as the top tech firms do. We need to understand its real value to us in all aspects of our lives. Blockchain technology can help us do that, enabling us to use our data proactively and improve our well-being. And while there are many areas where taking control of our data might improve our lives, there is one particularly promising ,,,, '

Learning Microwave Ovens

My microwave oven learns, say to cook a baked potato, but this takes it to a new dimension.  For possible industry applications.    See also my previous note on using microwave ovens for health data detection.  Use the tag below 'microwave'.  

Researchers develop 'learning' microwave ovens    by University of Amsterdam

In a publication in the Journal of Cleaner Production, Prof. Bob van der Zwaan of the Van 't Hoff Institute of Molecular Sciences presents the first example of a learning curve for microwave ovens, which follows a learning rate of around 20%. The paper discusses opportunities for possible microwave heating applications in households and industry that can contribute to sustainable development. Rapidly reducing prices could lead to a meaningful role of microwave technology in the energy transition.

Visualized Medical Record Analysis

Liked the idea, but also probably also requires a means of augmenting the visuals to emphasize key metadata, relationship to patterns detected by meta analyses, changes over time.

Visualization: U of T Researchers Develop System for Medical Records
U of T News   By Nina Haikara

Researchers at the University of Toronto (U of T) in Canada recently discussed their development of a curation-based approach for clinical text visualization. Using information from local clinics and other resources, the Doccurate visualization tool incorporated datasets of medical chart notes for each patient, with conditions cited throughout a chart visualized as steamgraphs, string-like graphs with droplets that expand upon recurrence of terms; each condition was labeled with a different color for quick recognition. U of T's Devin Singh said Doccurate stands apart from other medical visualization tools with its capability for customization. Said Singh, "It's communicated to me in a visual way, which helps link the mental models of physicians together, creating a holistic care team through visualization." U of T's Nicole Sultanum said the next step involves delivering a visual sense of narrative and progression as to how the patient has developed over time. ..

(Click through for complete graph)

Amazon and Medical Records

Amazon is Showing Healthcare is the Next Big Thing for Machine Learning

Amazon will save Healthcare industry $Billions via machine learning algorithm that extracts key data from patient records — it’s an EMR revolution by AI.
Michael K. Spencer

Amazon has had a health innovation stealth unit called 1492 for quite some time. We’re slowly starting to understand how sweeping its changes are going to be.

Recently we learned how Amazon will reportedly sell software that reads medical records. With ballooning healthcare costs anticipated in the next two decades globally, AI at the services of healthcare will be extremely important. It appears Amazon’s newest service uses machine learning to extract medical data from patient records. .... "

Medical IoT

Lab makes data sharing easier so medical IoT devices can be smarter

Medical PnP laboratory researchers are looking to save lives through smarter, more interoperable healthcare technology using open standards, medical expertise and testing equipment that simulates health conditions to encourage easier integration of IoT devices and new sharing apps that can expand their capabilities.

By Jon Gold,  Senior Writer, Network World

New Apple Watch can do an ECG in 30 Seconds

Watched the detailed Apple presentation of this yesterday.  Amazing the whole thing can now be put into  a Watch and they claim the (one lead) ECG can be requested and be received in thirty seconds.    An AI analysis is said to do the analysis.  Can be recorded and sent to you doctor on a PDF file.    Some elements of 'diagnosis' are involved.   The watch is expense at $400 plus. See in Wired 'Apple Watch Could do more Harm than Good'   https://www.wired.com/story/ecg-apple-watch/     Some physicians at the presentation had similar objections.    ... 

Precision Medicine Insight

Been a member of this Linkedin group for some time.   This announces that my blog will start to cover items in this space more often.    With particular emphasis on efficient therapeutic delivery aspects, and adding machine learning capabilities.  Comments welcome.   Other groups recommended?

Precision Medicine Insight   https://www.linkedin.com/groups/5180384

About this Group: 

Precision Medicine is an approach to discover and develop medicines, vaccines or routes of intervention (behavior, nutrition, etc.) that enable disease prevention and deliver superior therapeutic outcomes for patients, by integrating clinical, molecular (multi-omics including epigenetics), environmental and behavioral (Big Data) information to understand the biological basis of disease. 
This effort leads to better selection of disease targets and identification of patient populations that demonstrate improved clinical outcomes to novel preventive and therapeutic approaches.

In order to achieve this goal novel standards to harvest raw data (SOPs), processes, as well as architectures/analytics including new artificial intelligence, IoT and ICT technologies are instrumental to enable implementation and delivery of Precision Medicine 24/7 anywhere anytime in the world. 

Data in general will become the crude oil and currency of the future, while safety, security, ownership, privacy and so on constitute ethical challenges that need to be discussed alongside technical solutions to foster novel business and improve health care  ... " 

Electronic Health Records

In the Google Blog:

Deep Learning for Electronic Health Records
Posted by Alvin Rajkomar MD, Research Scientist and Eyal Oren PhD, Product Manager, Google AI

When patients get admitted to a hospital, they have many questions about what will happen next. When will I be able to go home? Will I get better? Will I have to come back to the hospital? Having precise answers to those questions helps doctors and nurses make care better, safer, and faster — if a patient’s health is deteriorating, doctors could be sent proactively to act before things get worse.

Predicting what will happen next is a natural application of machine learning. We wondered if the same types of machine learning that predict traffic during your commute or the next word in a translation from English to Spanish could be used for clinical predictions. For predictions to be useful in practice they should be, at least:

Scalable: Predictions should be straightforward to create for any important outcome and for different hospital systems. Since healthcare data is very complicated and requires much data wrangling, this requirement is not straightforward to satisfy.

Accurate: Predictions should alert clinicians to problems but not distract them with false alarms. With the widespread adoption of electronic health records, we set out to use that data to create more accurate prediction models. ... "

Cancer Algorithm and Game Theory

New heath care approaches.

Cancer algorithm uses game theory to Double Survival time

Using algorithms to monitor cancer evolution and apply game theory to their treatment has doubled the survival time of men with advanced prostate cancer    By Andy Coghlan in NewScientist

Approaching cancer treatment as a game has doubled the survival time of men with advanced prostate cancer. This achievement could mark the start of using game theory to target a range of cancers more cleverly.

“This approach is elegant and exciting, and shows real promise to delay treatment failure,” says Charles Swanton at the Francis Crick Institute in London.

People with cancer aren’t usually killed by their initial tumour, but by the rapidly evolving secondary tumours that occur once the disease …  "

Dealing with Health Data

Good and considerable piece.  Needs to be rethought before losing the complete power of medical data.

Finding a Healthier Approach to Managing Medical Data  By Samuel Greengard 
Communications of the ACM, Vol. 61 No. 5, Pages 31-33
10.1145/3193759

One of the formidable challenges healthcare providers face is putting medical data to maximum use. Somewhere between the quest to unlock the mysteries of medicine and design better treatments, therapies, and procedures, lies the real world of applying data and protecting patient privacy.

"Today, there are many barriers to putting data to work in the most effective way possible," observes Drew Harris, director of health policy and population health at Thomas Jefferson University's College of Population Health in Philadelphia, PA. "The goals of protecting patients and finding answers are frequently at odds."

It is a critical issue and one that will define the future of medicine. Medical advances are increasingly dependent on the analysis of enormous datasets—as well as data that extends beyond any one agency or enterprise. What's more, as connected healthcare devices flourish, at-home and remote monitoring blossoms and big data analytics advances at a staggering rate, the stakes—and the ability to use, misuse, and abuse confidential data grows significantly.

"Healthcare is at a very important crossroads. To move to a more value-based framework and one that rewards patient and doctor behavior, we need to have systems in place that manage data and protect individuals," says Ophir Frieder, professor of computer science and information processing at Georgetown University in Washington, D.C., and professor of biostatistics, bioinformatics, and biomathematics at the Georgetown University Medical Center.     .... "

Digital Twins for Personalized Medicine.

Reminds me in the human sense of agent based models.   And their use to address group behavior. Interesting, but not necessarily predictive.   The article quotes George Box, who said  'All Models are wrong, some are useful'.    How correct will the models be, how useful?   We expect personal accuracy from medicine.  Worth examining.

Digital Twins for personalized medicine: promising, with caveats    By Neil Raden  in SiliconAngle

" ... Digital Twins, a concept from the “industrial internet of things” or IIoT, is the discipline of devising highly capable simulation models, especially those that consume streaming data from sensors to anticipate maintenance issues, impending failure of components and improving performance. In terms of the degree of difficulty of modeling, Digital Twins for IIoT machines such as jet engines, oil wells and wind turbines are actually at the lower end as they employ detailed knowledge of the engineering of the objects. These Digital Twins may be quite complex, but because they are well-understood, the models are more likely to be useful.

Devising a simulation model of human behavior, such as the classical propensity models, is much more difficult because humans are so unpredictable and engineering approaches obviously don’t apply. In effect, these models may do a good job of predicting propensity based (on) many criteria, but at any given moment for any given individual, their predictive capability is quite low.  .... " 

Using Knowledge Studio

For the developer, a video showing how to use Knowledge Studio for a classic form of machine learning application.  Development-technical, but relatively straight forward example

Use Watson Knowledge Studio to build a custom machine learning model in the medical domain

About this webcast

One of the key benefits of building a machine learning annotator is the ability to train Watson in a complex domain such as medicine. Learn the methodology, standard practices, and guiderails on how to go about building an effective ML model. Steps include data understanding, type system building, pre-annotation, and deployment to WDS. After this session, you will have an acute understanding of what goes behind building an effective ML model. ... "

Spotting Viruses with Machine Learning

A further indication of the value of pattern recognition ...

Machine Learning Spots Treasure Trove of Elusive Viruses 
In Nature  by Amy Maxmen

Researchers have applied artificial intelligence to the discovery of nearly 6,000 previously unknown species of virus. Simon Roux at the U.S. Department of Energy's Joint Genome Institute trained computers to identify viral genetic sequences from the Inoviridae family. He gave a machine-learning algorithm two sets of data, including 805 genomic sequences from known Inoviridae viruses, and about 2,000 sequences from bacteria and other types of virus, so the program could learn to differentiate between them. The model was then fed metagenomic datasets, and the computer retrieved more than 10,000 Inoviridae genomes, clustering them into groups indicative of different species. A separate study conducted at the University of Sao Paulo in Brazil used machine learning to identify viruses in compost piles at the city's zoo, by programming an algorithm to look for a few distinguishing characteristics of virus genomes. Following training, the computer recovered several genomes that appeared to be new. .... " 

High Quality Images from Limited data

Always looking for more information from less data.  You may just not have enough data, or it may take more effort, time, cost or exposure to get it. 

New artificial intelligence technique dramatically improves the quality of medical imaging

Source:   Massachusetts General Hospital in ScienceDaily.

Researchers have developed a new technique based on artificial intelligence and machine learning that should enable clinicians to acquire high-quality images from limited data. ... "

Blood Pressure Sensing from a Smart Phone

Was involved in defense department look at how to do inobtrusive, non mechanical blood pressure sensng.   Was considered impossible at one time.   The method described does not require complex calibration. Availability not mentioned. Includes video.

New Smartphone Sensor Checks Your Blood Pressure
It’s more convenient than a cuff   and could help patients monitor hypertension at home
By Emily Waltz in IEEE Spectrum

For years, scores of engineers have been trying to develop a more unobtrusive, convenient device for blood pressure monitoring. Now, researchers at Michigan State University and University of Maryland appear to have succeeded.

In a paper published today in Science Translational Medicine, the researchers described a prototype blood pressure sensor that can be incorporated into a smartphone, and requires only the press of a fingertip.  

The convenient device could encourage people to check their blood pressure more often, allowing them to catch hypertension—persistently high blood pressure—sooner, says Ramakrishna Mukkamala, a biomedical engineer at Michigan State, in East Lansing, who led the study.   ... " 

Computer Vision Assessing Cardiovascular Risk

From Google Research, an interesting application.  With considerable detail about the experimental approach.   Instructive.   Note its about risk rather than diagnosis.   Examples of retinal images.  Would like to see more medical /statistical commentary on the results.

Assessing Cardiovascular Risk Factors with Computer Vision
Posted by Lily Peng MD PhD, Product Manager, Google Brain Team 

Heart attacks, strokes and other cardiovascular (CV) diseases continue to be among the top public health issues. Assessing this risk is critical first step toward reducing the likelihood that a patient suffers a CV event in the future. To do this assessment, doctors take into account a variety of risk factors — some genetic (like age and sex), some with lifestyle components (like smoking and blood pressure). While most of these factors can be obtained by simply asking the patient, others factors, like cholesterol, require a blood draw. Doctors also take into account whether or not a patient has another disease, such as diabetes, which is associated with significantly increased risk of CV events. 

Recently, we’ve seen many examples [1–4] of how deep learning techniques can help to increase the accuracy of diagnoses for medical imaging, especially for diabetic eye disease. In “Prediction of Cardiovascular Risk Factors from Retinal Fundus Photographs via Deep Learning,” published in Nature Biomedical Engineering, we show that in addition to detecting eye disease, images of the eye can very accurately predict other indicators of CV health. This discovery is particularly exciting because it suggests we might discover even more ways to diagnose health issues from retinal images.  .... " 

Recalculating Patterns in Real Time

Intriguing approach for constantly changing time series data.

Novel Algorithm Enables Statistical Analysis of Time Series Data 

MIT News  By Sara Cody

Researchers at the Massachusetts Institute of Technology (MIT) have developed state-space multitaper time-frequency analysis (SS-MT), a unique algorithm they say delivers time series dataset analysis in real time. The team notes SS-MT enables scientists to work in a more informed manner with large, nonstationary datasets so they can not only measure the fluid properties of data but also make formal statistical comparisons between arbitrary data segments. "The algorithm functions similarly to the way a [global-positioning system] calculates your route when driving," says MIT professor Emery Brown. The team tested SS-MT by first analyzing electroencephalogram readings from patients receiving general anesthesia for surgery. The program produced a de-noised spectrogram defining changes in power across frequencies over time, and the researchers also applied SS-MT's inference paradigm to compare different levels of unconsciousness in terms of the differences in the spectral properties of these behavioral states. "The SS-MT analysis produces cleaner, sharper spectrograms," Brown says.  .... "

Humans vs AI in Healthcare Vision

Google Brain chief Jeff Dean: As AI beats humans in computer vision, healthcare will never be the same

SiliconANGLE by Gina Smith

Just five years ago, artificial intelligence-enabled computers could barely recognize images fed to them, much less analyze them anything like people can. But suddenly, they’ve turned the tables.

“In 2011 their error rate was 26 percent,” says Jeff Dean, chief of the Google Brain project, which along with other tech giants has helped lead a recent revolution in image recognition as well as speech recognition and self-driving cars. Now, he says, computers’ ability to view and analyze images (pictured) exceeds what human eyes can do.

“If you’d have told me that would be a possible just a few years ago, I would’ve never believed you,” Dean said during an appearance at a research event in Heidelberg, Germany. But thanks to AI-enabled computer vision advances, computers “can now see … and that has opened our eyes (about) what is possible.”.... ' 

Analytics to Combat Opiod Epidemic

Becoming an epidemic in our area.  Good thoughts here, but would have liked more in the area of pattern detection, predictive and prescriptive applications.  Though once you did more accurate data gathering,  you would get more opportunity for intelligent applications. Someone doing that specifically in this space, contact me, willing to support ?

How law enforcement can use analytics to combat the opioid epidemic   
By David Kennedy on SAS Voices 

A steady drumbeat of news coverage makes one thing clear: Opioid abuse is rising and has reached epidemic levels throughout our country. Overdoses from the diversion and abuse of prescription opioids are one cause of the surge in deaths. Overdoses from heroin and other illicit synthetic opioids (such as heroin, fentanyl and carfentanil) are an even greater issue, resulting in approximately twice as many fatalities as prescription opioids. Law makers, health departments and the medical community are all searching for ways to slow the volume and diversion of prescription drugs, but law enforcement officers are at the front lines of the battle.

Domestic violence calls, emergency medical calls, narcotics investigations and patrol-related activity can all involve an officer encountering someone overdosing from opioids. Today, many officers are put in the role of drug counselor and medical worker as much as they are enforcement officer. We must understand what officers are experiencing, what resources they need to more effectively help those in need -- and the future threats they may face. ... "