Big Surprise - Big Data - AI

Bloggen har varit ute med grävspaden igen och hittat en Surprise, en Big Surprise t.o.m.
Det handlar om AI, Artificiell Intelligens. Att få datorer via s.k Machine Learning sortera,katalogisera och tolka / förstå det som kallas Big Data. Alltså den samlade information man får ut av att använda sig av digital teknik. Forskare som använder PHI´s teknik Digital Holomonitoring Microscopy förkortat DHM får massiv information vid användande av HoloMonitor. Men det är enbart en bråkdel av den informationen forskarna förstår och använder sig av.
Jag får citera mig själv från ett inlägg jag skrev 2017 beträffande machine learning (då jag fasen inte hittar ursprungscitatet)
"Jag tror mig ha läst nånstans att vi enbart förstår och använder runt 10% av all den info DHM lämnar."
Tekniken medger alltså det som kallas Big Data, en informationsbank som man ännu inte listat ut hur att använda sig av.
VD Egelberg var redan 2015 klar över vad som komma skall:

Big data

Modern information technology has given cell biologists an untapped tool to make sense of cellular interactions in a large cell population.
We all use the technology when we use Google, but for a completely different purpose.
Google uses what is known as Big Data technology.
This technology applies computer algorithms to help us select relevant information from a bewildering amount of information that our brains simple cannot otherwise cope with.

A future cell laboratory

Imaged-based time-lapse cytometers, as our HoloMonitor, creates huge amounts of information on how cells interact and multiply over long time periods. Com­bi­ned with Big Data technology, such novel instru­menta­tion will provide cell bio­lo­gists and the pharmaceutical industry with a game changing new tool to better understand complex cellular interactions and through this hopefully cure cancer.
This is why the pharmaceutical industry is excited about HoloMonitor^® – it fundamentally changes cell biologists’ view on cell biology.

PHI har 2 partners som börjat närma sig att få ut mer info av tekniken. Det är Ed Luther m kollegor på Northeastern University och Robert Judson på University of Californa.
Från PHI`s hemsida en notis från 2016

Big data analysis

At the recent scientific conference on Engineering and Physical Sciences in Oncology organized by the American Association for Cancer Research (AACR), scientists from Northeastern University presented a big data analysis method based on a modified Kolmogorov-Smirnov test.
The scientists have used the newly developed method to compare time series data of drug treated cells with untreated cells. The work was spear­headed by Ed Luther who manages the Holographic Imaging Cytometry Program of Excellence at Northeastern University.

Sen har vi Robert Judson som 2017 var och sniffade på området med egenuttänkt teknik via machine learning.
Det utmynnade i denna forskningsrapport. Ur den klistrar jag in:
Abstract:
Digital holographic microscopy permits live and label-free visualization of adherent cells.
Here we report the application of this approach for high accuracy kinetic quantitativ cytometry.
We identify twenty-six label-free optical and morphological features that are biologically
independent. When used as a basis for machine learning, these features allow blind single cell
classification with up to 95% accuracy. We present methods to control for inherent holographic
noise, thereby establishing a set of reliable quantitative features. Together, these contributions
permit continuous digital holographic cytometry for three or more days. Applying our approach
to human melanoma cells treated with a panel of cancer therapeutics, we can track the response
of each cell, simultaneously classifying multiple behaviors such as cell cycle length, motility,
apoptosis, senescence, and heterogeneity of response to each therapeutic. Importantly, we
demonstrate relationships between these phenotypes over time. This work thus provides an
experimental and computational roadmap for low cost live-cell imaging and kinetic classification
of heterogeneous adherent cell populations.

Holograms were captured using a HoloMonitor M4 DHC system and acquired every hour to observe all cell divisions in the non-treated conditions. The holograms were processed computationally using HStudio to produce an intensity image representing a quantitative map of light-wave phase shifts.
We next tested the capacity of our platform for deep screening by treating A375 human melanoma cells with a panel of well-characterized kinase inhibitors known to have toxic, growth-arresting or negligible effects.
By first establishing biologically homogeneous mammalian cell populations, we demonstrate that an extended set of features are amenable to machine-learning driven phenotypic profiling, increasing the accuracy of classification from a variable range of 11-89% to a more reliable range of 82-100%.
Further, we developed methods for standardizing hologram quality for both experimental setup and long-term time-lapse imaging.
The basic technical set-up for DHM is relatively simple as compared to fluorescent microscopy and is consequently a more affordable option for phenotypic profiling. The system used in this study, the HoloMonitor M4, can be assembled inside a standard mammalian tissue culture incubator. DHM is inherently label-free and non-cytotoxic, allowing for long-term imaging of cell populations and tracking of individual cells within a population. As this analysis is non-endpoint, the cells can be used for a variety of purposes once the analysis is complete – for example, molecular verification of cell state as we did here, further propagation, or injection into a mouse.
We foresee a variety of uses for this platform.

"These analyses demonstrate that DHC ( Digital Holographic Contouring) is a sufficient platform for reliable label-freesingle cell classification using machine-learning-based phenotypic profiling."
-----------------------------------------------------------------------------------------------------------------------------
Ed Luther och Robert Judson m forskarkollegor är snillen inom cellforskning, men kanske inte lika framstående inom datateknik. Att bygga dataprogram som gör jobbet istället för att manuellt sitta och lista ut all information. Här behövs mer än Human Intelligence ---> Artificial Intelligence, AI.
Då krävs det datasnillen som kan konstruera program via s.k algoritmer.
Och då kära läsare kommer bloggen in på the Big Surprise. 

Bloggen har nämligen hittat ett datasnilleföretag som enbart jobbar med att bygga programvaror som tolkar information inom medicinforskningsområdet.
Och som har/haft uppdraget att bygga programvara kring HoloMonitor specifikt.
Det handlar om det Estländska företaget Matimex Group.
Såhär presenterar de sin verksamhet:

AI (Artificial Intelligence) in Medicine

Matimex Group is a software development company, based in Estonia. We are creating medical services and web application that analyse big data sets of Medical documents.
Deep machine learning we use, can quickly and cost-effectively analyze lots of information from structured claim data to discovery notes and diaries, which are unstructured data.

We are an Estonian company working as outsourcing contractor for medical AI systems. 
Estonia is one of the world’s leading innovative countries, many IT startups were born in Estonia, such as Skype, Playtech, TransferWise, Taxify, and many-many more. 
We are working in the united Estonian IT eco-system and provide our customers wide range of high effective development of AI enabled medical systems. 
Matimex Group was founded in 2002 by two IT professionals, Savva Motovilov (IA/UI) and Dmitry Verkhovsky (CTO) as a web-development production bureau. The company had contracts for medical infosystems outsourcing development. Typical projects were based on EHR (electronical healthcare records) systems and realized on LAMP (Linux-Apache-MySQL-PHP) stack. 
In the next 5 years we improved the architecture and application design to satisfy customers growing requirements in stability and redundancy.  

Our Mission

Do innovative medicine IT systems through our technologies in Artificial Intelligence.

Our Philosophy

Developing breakingthrough technologies and application by implementing the most disruptive emerging trends. Inventing the processes based on providing algorithms, developing APIs (Application Programming Interfaces), combining big data analysing and DML (Deep Machine Learing) training activities. Discovering (AI) Artificial Intelligence technologies and bringing new ideas.

På deras hemsida under rubrik Projects hittar vi följande:

Projects

Keyframes analyzer for Microscopy videostream [2018]

Digital holographic microscopy video analyzing. Searching keyframes and segmenting cell divisions of unlabeled JIMT-1 breast cancer cells.

The JIMT-1 cells were seeded on a Ibidi[TM] (Ibidi Gmbh Martinsried,Germany), one well slide and left to incubate for one day (5% CO2, 90% humidity, 37 degrees C).
The Ibidi slide was placed on a Ibidi heating plate, set to 37 degrees C, and onto the stage of the HoloMonitor M3 (Phase Holographic Imaging AB, Lund Sweden).

Just denna text/forskning är från 2017 och utförd av forskaren Birgit Jannicke,närstående PHI, och handlar om bröstcancerforskning. Här kan man se ett utdrag med film där bilderna ovanför kommer från.

Min kommentar
Att Matimex använt sig av just denna forskning kan bero på att det finns massvis med Big Data att hämta ur den.
Vi får årtalet 2018 angivet vilket borde indikera att Matimex då startade sitt uppdrag.
Sen kommer några intressanta frågeställningar.
Vem har gett Matimex uppdraget att utveckla/bygga algoritmer/dataprogram byggt på denna forskningsrapport?
Som vi kan se ur deras presentation åtar de sig uppdrag (outsorcing contractor) på beställning.
Det lär kosta flis så någon aktör vill väldigt gärna få tillgång till ALL information HoloMonitortekniken kan ge.
Är det en befintlig kund som ligger bakom? Kanske det där mystiska Big Pharma bolaget vi ännu inte vet namnet på? "SWITZERLAND Major pharmaceutical company, 3 units"
Eller är det BioSpherix som har planer på att maxa marknadsföringen med "tillbehöret" programvaran HoloMonitor AI när de väl kör igång säljeriet?
Någon annan?
Sen att Matimex först nu visar upp uppdraget kan det tyda på att det är klart och programvaran redo att användas?
Och hur mycket information har datasnillena lyckats få fram genom programvaran?
Såpass att tidigare forskning byggd på HoloMonitor kan ge fler svar? Lösningar?
Vad innebär denna programvara för framtida forskning?
Man kan spekulera en hel del kring ovanstående (vilket som de flesta av er vet är bloggens favoritämne 😎)
Att detta är intressant tror jag de flesta förstår. Vi har nånting spännande att invänta. En Big Surprise?

                                               Vid grävspaden the99
Ps. Informationen om Matimex/HoloMonitor blev tillgänglig på nätet för 2 dagar sen. Ds