BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Replacing pathologists with next generation diagnostics: use of ar
 tificial intelligence to analyse image and DNA sequencing data - Dr Elizab
 eth Sollieux (Dept of Pathology)\, Dr Matt Thorpe (DAMTP) and Oliver Crook
  (MRC Biostatistics Unit/DAMTP)
DTSTART:20190130T140000Z
DTEND:20190130T150000Z
UID:TALK116296@talks.cam.ac.uk
CONTACT:46925
DESCRIPTION:Abstract:\nTo date\, tissue-based diagnostics have been inhere
 ntly subjective. To make the process more robust\, we are developing an ar
 tificial intelligence solution based on microscopic image analysis. Ultima
 tely\, we wish to combine this with a molecular analysis that measures imm
 une responses in the same tissue\, to provide a completely novel\, holisti
 c approach able to predict diagnosis and/ or prognosis in a wide range of 
 conditions mediated or modulated by the immune system. In this exciting pr
 oject\, we will develop an advanced mathematical classification of a relat
 ively heterogeneous disease.  \n\nThe model we will use is that of coeliac
  disease (CD)\, a relatively common condition\, caused by a damaging immun
 e response to the small intestine\, with effects ranging from no symptoms\
 , through anaemia to severe intestinal symptoms\, and with complications i
 ncluding bone thinning\, infertility and\, rarely\, lymphoma and duodenal 
 cancer.  Assessing whether a patient has CD relies primarily on “gold st
 andard” duodenal biopsy examination by a pathologist\, an unavoidably su
 bjective process with poor inter-observer concordance (up to 25% cases) an
 d a suspicion that a significant number of cases are missed.  Furthermore\
 , some biopsies appear to show a disease continuum from normal to severe. 
 \n\nWe will discuss image analysis work undertaken on a subset of our coho
 rt of carefully clinicopathologically annotated\, high resolution digitise
 d images of 500 duodenal biopsy samples (250 coeliac disease\; 200 normal\
 ; 50 equivocal).  We will also explain some of our future image analysis p
 lans.  Finally we will consider how we might undertake a combined integrat
 ive analysis of digital images\, together with a novel molecular analysis 
 of the activity of the immune system\, the algorithm for which was develop
 ed in the Soilleux laboratory.  We believe that the algorithms generated d
 uring this project hold great promise in a broad range of areas\, includin
 g diagnostics. There is relatively little work on joint analysis of sequen
 ce and imaging data and the machine learning methodologies we develop will
  ultimately be useful for a range of immune-modulated disease.\n
LOCATION:MR4\, Centre for Mathematical Sciences\, Wilberforce Road\, Cambr
 idge
END:VEVENT
END:VCALENDAR