Computational Pathology Workshop July 8 2014

36

description

 

Transcript of Computational Pathology Workshop July 8 2014

Page 1: Computational Pathology Workshop July 8 2014
Page 2: Computational Pathology Workshop July 8 2014

Computa(onal  Pathology:  Research  

Joel  Saltz  MD,  PhD  Chair  Biomedical  Informa(cs  Stony  

Brook  University  Associate  Director  for  Informa(cs,  

Stony  Brook  Cancer  Center  

Page 3: Computational Pathology Workshop July 8 2014

Computa(onal  Pathology  Research  

•  Computa(onal  Science  –  Context  • High  Dimensional  Fused  Informa(cs  •  Internet  of  People  and  Things  

Page 4: Computational Pathology Workshop July 8 2014

Computa(onal  Science  

Page 5: Computational Pathology Workshop July 8 2014

Detect and track changes in data during production

Invert data for reservoir properties Detect and track reservoir changes

Assimilate data & reservoir properties into

the evolving reservoir model Use simulation and optimization to guide future production

Example:  Oil  Field  Management  –  Joint  ITR  with  Mary  Wheeler,  Paul  Stoffa  

Page 6: Computational Pathology Workshop July 8 2014

Coupled  Ground  Water  and  Surface  Water  Simula(ons  

Multiple codes -- e.g. fluid code, contaminant transport code Different space and time scales Data from a given fluid code run is used in different contaminant transport code scenarios

Page 7: Computational Pathology Workshop July 8 2014

Pete Beckman – Workshop on Big Data and Extreme Scale Computing

Page 8: Computational Pathology Workshop July 8 2014

Titan  –  Peak  Speed  30,000,000,000,000,000  floa(ng  point  opera(ons  per  second!  

Pete Beckman – Workshop on Big Data and Extreme Scale Computing

Page 9: Computational Pathology Workshop July 8 2014

Computa(onal  Pathology:  High  Dimensional  Fused-­‐Informa(cs  

•  Anatomic/func(onal  characteriza(on  at  fine  and  gross  level    

•  Integrate  of  anatomic/func(onal  characteriza(on,  mul(ple  types  of  “omic”  informa(on,  outcome  

•  Predict  treatment  outcome,  select,  monitor  treatments  

•  Integrated  analysis  and  presenta(on  of  observa(ons,  features  analy(cal  results  –  human  and  machine  generated  

Ex-­‐vivo  Imaging  

Pa.ent    Outcome  

In  vivo  imaging  

“Omic”  Data              

Page 10: Computational Pathology Workshop July 8 2014

Correlating Imaging Phenotypes with Genomic Signatures: Scientific Opportunities

(Imaging Genomics Workshop NCI June 2013)

Clinical Approach and Use •  Development of imaging+analysis methods to

characterize heterogeneity •  within a tumor at one time point •  evolution over time •  among different tumor types

•  Development of imaging metrics that: •  can predict and detect emergence of resistance? •  correlates with genomic heterogeneity? •  correlates with habitat heterogeneity? •  can identify more homogeneous sub-types

Page 11: Computational Pathology Workshop July 8 2014

Tumor Heterogeneity

Marusyk 2012

Page 12: Computational Pathology Workshop July 8 2014

Pathology  Analy(cal  Imaging  

•  Provide   rich   informa(on   about   morphological   and  func(onal  characteris(cs  

•  Image  analysis,  feature  extrac(on  on  mul(ple  scales  •  Spa(ally  mapped  “omics”  •  Mul(ple  microscopy  modali(es  

Glass Slides Scanning Whole Slide Images Image Analysis

Page 13: Computational Pathology Workshop July 8 2014

•  Quantitative Feature Analysis in Pathology: Emory In Silico Center for Brain Tumor Research (PI = Dan Brat, PD= Joel Saltz)

•  NLM/NCI: Integrative Analysis/Digital Pathology R01LM011119, R01LM009239 (Dual PIs Joel Saltz, David Foran)

•  New - NCI: 1U24CA180924-01A1 Tools to Analyze Morphology and Spatially Mapped Molecular Data (PI=Saltz)

Page 14: Computational Pathology Workshop July 8 2014

Direct Study of Relationship Between vs

Lee Cooper, Carlos Moreno

Page 15: Computational Pathology Workshop July 8 2014

Clustering identifies three morphological groups •  Analyzed 200 million nuclei from 162 TCGA GBMs (462 slides)

•  Named for functions of associated genes: Cell Cycle (CC), Chromatin Modification (CM), Protein Biosynthesis (PB)

•  Prognostically-significant (logrank p=4.5e-4)

Feat

ure

Indi

ces

CC CM PB

10

20

30

40

500 500 1000 1500 2000 2500 3000

0

0.2

0.4

0.6

0.8

1

Days

Surv

ival

CCCMPB

Page 16: Computational Pathology Workshop July 8 2014

Associations

Page 17: Computational Pathology Workshop July 8 2014

Gene Expression Correlates of GBM with High Oligo-Astro Ratio

Oligo Related Genes Myelin Basic Protein Proteolipoprotein HoxD1 Nuclear features most Associated with Oligo Signature Genes: Circularity (high) Eccentricity (low)

Page 18: Computational Pathology Workshop July 8 2014

Microenvironment  and  Master  Regulators  

•  Extent  of  Necrosis  Related  Expression  of  Master  Regulators  of  the  Mesenchymal  Transi(on  

Necrosis and C/EBP-β

Page 19: Computational Pathology Workshop July 8 2014

Computa(on  and  Data  Management:  Requirements  and  Challenges  

•  Explosion  of  derived  data  –  105x105    pixels  per  image  –  1  million  objects  per  image  –  Hundreds  to  thousands  of  images  per  study  

•  High  computa(onal  complexity  –  Image  analysis,  feature  extrac(on,  machine  learning  pipelines  

–  Spa(al  queries  involve  heavy  duty  geometric  computa(ons  

Page 20: Computational Pathology Workshop July 8 2014

Projec(on  –  2025      

•  100K  –  1M  pathology  slides/hospital/year  •  2GB  compressed  per  slide  •  1-­‐10  slides  used  for  Pathologist  computer  aided  diagnosis  

•  100-­‐10K  slides  used  in  hospital  Quality  control  •  Groups  of  100K+  slides  used  for  clinical  research  studies  -­‐-­‐  Combined  with  molecular,  outcome  data  

Page 21: Computational Pathology Workshop July 8 2014

HPC:  Tools  for  Image  Analysis,  Feature  Extrac.on,    Machine  Learning  Pipelines  

Page 22: Computational Pathology Workshop July 8 2014

Large  Scale  Data  Management  

Ø Data   model   capturing   mul(-­‐faceted   informa(on  including   markups,   annota(ons,   algorithm  provenance,  specimen,  etc.  

Ø Support  for  complex  rela(onships  and  spa(al  query:  mul(-­‐level   granulari(es,   rela(onships   between  markups   and   annota(ons,   spa(al   and   nested  rela(onships  

Ø Highly  op(mized  spa(al  query  and  analyses  Ø Implemented   in   a   variety   of   ways   including  op(mized  CPU/GPU,    Hadoop/HDFS  and    IBM  DB2    

PAIS Database Ø Implemented with IBM DB2 for large scale pathology image metadata (~million markups per slide) Ø Represented by a complex data model capturing multi-faceted information including markups, annotations, algorithm provenance, specimen, etc. Ø Support for complex relationships and spatial query: multi-level granularities, relationships between markups and annotations, spatial and nested relationships Ø Support for high-level data statistical analysis

Page 23: Computational Pathology Workshop July 8 2014

Spa(al  Centric  –  Pathology  Imaging  “GIS”  Point  query:  human  marked  point    inside  a  nucleus  

.  

Window  query:  return  markups    contained  in  a  rectangle  

Spa.al  join  query:  algorithm    valida(on/comparison  

Containment  query:  nuclear  feature  aggrega(on  in  tumor  regions  

Fusheng Wang

Page 24: Computational Pathology Workshop July 8 2014

MICCAI 2014 BRAIN TUMOR

Classification and Segmentation Challenges

TCGA  

TCIA  

IMAGING    CHALLENGE  

DIGITAL  PATHOLOGY  CHALLENGE  

Phase  1:  Training  June  20  -­‐  July  31  

Phase  2:  Leader  Board  Aug  1  -­‐  Aug  29  

Phase  3:  Test  Sept  8  -­‐  Sept  12  

For  more  informa+on  about  these  challenges  and  a  related  workshop    on  September  14,  2014  at  MICCAI  in  Boston,  see:  cancerimagingarchive.net  

MICCAI:  Medical  Image  Compu.ng  and  Computer  Aided  Interven.ons  -­‐  MICCAI2014.org  TCGA:    The  Cancer  Genome  Atlas  -­‐  cancergenome.nih.gov  TCIA:  The  Cancer  Image  Archive  -­‐  cancerimagingarchive.net  

Page 25: Computational Pathology Workshop July 8 2014

Digital  Pathology/Brain  Tumor  Image  Segmenta(on  (BRATS)  

•  Use  data  currently  available  through  data  archive  resources  of  the  Na(onal  Ins(tutes  of  Health  (NIH),  namely,  the  Cancer  Genome  Atlas  (TCGA)  and  the  Cancer  Image  Archive  (TCIA)    

•  Digital  Pathology  challenge  will  use  digital  slides  related  to  pa(ents  whose  genomics  data  are  available  from  TCGA.  Similarly,  BRATS  2014  Challenge  will  use  clinical  MRI  image  data,  also  from  the  TCGA  study  subjects.  

•  Coordinated  Pathology/Radiology  2015  challenge    –  feature  selec.on  and  sta.s.cal/machine  learning  algorithms  to  leverage  Radiology,  Pathology  and  “omic”  features  to  predict  outcome,  response  to  treatment  

Page 26: Computational Pathology Workshop July 8 2014

Computa(onal  Pathology:  Popula(ons  

Page 27: Computational Pathology Workshop July 8 2014

Suffolk County PPS IT Architecture Suffolk  County  

Providers  

Suffolk  county  PPS  Master  Pa.ent  Index  (MPI)  

Suffolk  county  PPS  Health  Informa.on  Exchange  (HIE)  

E-­‐HNLI  RHIO  (HIE)  

Suffolk  County  PPS  Pa.ent  Portal      

Stony  Brook  Medicine  

     

Suffolk  County  Big  Data  Plaaorm                

Suffolk  County  PPS  Popula.on  Management  Tools    

EMRs  or  clinical  Informa.on  System  EMRs  or  clinical  Informa.on  System  

eForms   Pa(ent  Wellness  

Alerts   Mobile  Monitoring  

Pa(ent  Educa(on  

Clinical  Records  

Collabora(on  

Predic(ve  Analy(cs   Event  Engine   Structured  Data   Financial  Data   Legacy  Data  

Machine  Learning   NLP   Unstructured  Data   Wearables  Data   Social  Data  

Anomaly  Detec(on   Rules   Device  Data   HL7/CCD   Open  Data  

Clinical  Data  for  P

a.en

t  Care  

Jim Murry CIO, Charles Boisey

Page 28: Computational Pathology Workshop July 8 2014

Suffolk  PPS  Organiza(onal  Structure  for  exchange  of  clinical  data  and  alerts  for  pa(ent  visits  

through  e-­‐HNLI  

         

Stony  Brook  Medicine  

Suffolk  PPS  HIE  (SB  Clinical  

Network  IPA,  LLC)  

Health  Systems  

Hospitals  

Community  Health  Centers  

Behavioral  Healthcare  Providers  

Skilled  Nursing  Facili.es   CHHA’s/  

LTHHC  

Physician  Groups  

Health  Homes  

Community-­‐Based  Agencies  

Pharmacies  

Those not part of the Stony Brook Medicine

Network

Other  Healthcare  Providers  

Develop-­‐mental  Disability  Providers  

6

Suffolk  county  RHIO  (e-­‐HNLI)  

Jim Murry CIO, Charles Boisey

Page 29: Computational Pathology Workshop July 8 2014

The Internet of People and Things

•  Distributed mHealth devices, sensors, point of care devices, EHRs computers and databases

•  Collections of interacting services •  Ubiquitous access to all clinical, laboratory, sensor,

radiology, pathology, treatment data •  Iteratively scan patient information to evaluate

interventions •  Aggregate and iterative mine patient information to

evaluate how to optimize treatment •  Predictive/interactive analytics that anticipate

problems and launch preventive measures •  QC/QA on data and process

Page 30: Computational Pathology Workshop July 8 2014

Minimize Surprise

•  Evaluate, track, quantify progression of known disease states

•  Track, evaluate risk factors and carry out diagnostic screenings where risk factors are significant

•  Active learning to formulate correct questions to ask •  When unanticipated catastrophic event occurs, or

disease is first found in advanced state carry out systematic retrospective population study –  Identify what was different about “surprise”

patients and unaffected cohorts

Page 31: Computational Pathology Workshop July 8 2014

Our work at Emory: Find hot spots in readmissions within 30 days –  Integrative analysis - crucial lab data role - to

characterize co-morbidities and clinical course –  What fraction of patients with a given principal

diagnosis will be readmitted within 30 days? –  What fraction of patients with a given set of

diseases will be readmitted within 30 days? –  How does severity and time course of co-

morbidities affect readmissions?

EMR Data Analytics: Tools for Clinical Phenotyping and Population Health

Page 32: Computational Pathology Workshop July 8 2014

Johns Hopkins Medical Institutions

Department of Pathology Johns Hopkins

(1999) Joel Saltz MD, PhD – Director Pathology Informatics

Jim Nichols, MD -- Assistant Professor JHU and head of POCT Program

Merwyn Taylor, PhD -- Instructor, Informatics Division, Dept of Pathology, JHU

Laboratory Without Walls

Page 33: Computational Pathology Workshop July 8 2014

Johns Hopkins Medical Institutions

POCT Anywhere ●  Provide patients with up-to-date clinical data,

interpretations of clinical data and health related educational materials *  Integrated archive of patient clinical information,

education materials used by patients, families and health care providers

●  Maintain collection of medical information gathered at patient’s home, in clinics and during hospitalizations Alert clinicians about abnormal values, non-compliance

●  Interactive monitoring of POC device

Page 34: Computational Pathology Workshop July 8 2014

Where  Does  Pathology  Fit  In?  

•  Capture  and  analysis  of  laboratory  data  is  Pathology  •  Sensor  data  can  be  thought  of  as  generalized  lab  data  

•  Clinical  Pathology:  data  quality,    process  control,  sta(s(cal  analyses,    analy(c  vs  biological  varia(on  

•  Predic(ons  improved  by  including  novel  tests  –  reduc(on  of    “omics”  to  rou(ne  clinical  tes(ng  

•  Pharmacogenomics  is  just  the  beginning  ….  

Page 35: Computational Pathology Workshop July 8 2014

Where  does  Computa(onal  Pathology  Fit  In?  

•  Machine  learning  and  predic(ve  analy(cs  algorithms  applied  to  popula(on  health  

•  Context  sensi(ve  modeling  of  how  integrated  data  from  mul(ple  sources  influences  probability  distribu(ons  associated  with  different  health  condi(ons  

•  Applied  popula(on  “omics”  •  Integra(on  and  analysis  of  data  from  pa(ent  sensors  •  Integra(on  and  analysis  of  spa(al  data  sources    

Page 36: Computational Pathology Workshop July 8 2014

Thanks!