Skip to main content
The NCI Community Hub will be retiring in May 2024. For more information please visit the NCIHub Retirement Page:https://ncihub.cancer.gov/groups/ncihubshutdown/overview
close
eeDAP studies Logo
  • Discoverability Visible
  • Join Policy Restricted
  • Created 13 Jan 2017

eeDAP studies

►

Wiki

Introducing CSHL Team

Version 1
by (unknown)
Version 2
by (unknown)

Deletions or items before changed

Additions or items after changed

1 -
Greetings, and thanks for the invite. I joined the nciphub group.
+
Team:
2
3 -
I am interested in image analytics using both classical and modern machine learning methods for purposes of analyzing microscopic images of histological sections. My interested is based in my work analyzing a large (petabyte) data set of light-microscopic histological images I have gathered to map mouse brain connectivity at CSHL. I am interested in the related problem of image-based diagnostics in cancer, and have recently looked collaboratively (with colleagues at IIT Madras and at Northwell) at mitotic figure counting, the subject of your proposed study. I am also interested in the scientific question of comparing human and machine performance in this area, as a fundamental neuroscientific problem.
+
Partha P Mitra (CSHL) http://mouse.brainarchitecture.org
4
5 -
All best [[br]]
+
Adam Kepecs (CSHL) http://kepecslab.cshl.edu/
6 -
Partha Mitra [[br]]
+
7 -
Crick-Clay Professor of Biomathematics [[br]]
+
TBD Postdoc (CSHL)
8 -
Cold Spring Harbor Laboratory, USA. [[br]]
+
9 -
H N Mahabala Visiting Chair Professor, IIT Madras, India. [[br]]
+
James Crawford (Northwell) http://www.feinsteininstitute.org/our-researchers/james-m-crawford-md-phd/
10 -
Senior Visiting Researcher, RIKEN BSI, Japan. [[br]]
 +
11 +
TBD Pathology resident (Northwell)
12 +
13 +
Also: In collaboration with IIT Madras team also on this wiki
14 +
15 +
Objectives:
16 +
17 +
An important emphasis of modern machine learning is to discover “implicit knowledge” from labeled examples or via reinforcement learning. In the case of automating pathology diagnoses, which is the context of this study, the implicit knowledge in question is embedded in the pathologist, and includes the capabilities of the visual system, skilled judgment of microscopic histopathology images. and background biomedical knowledge.
18 +
19 +
One of our objectives is to examine this implicit knowledge by psychophysical experiments in which the participants are skilled pathologists judging histopathological sections under a microscope. These judgments will be compared to the performance of machine learning algorithms trained to mimic the performance of the pathologists but acting on co-registered whole slide image data.
20 +
21 +
By comparing and contrasting the performance of the pathologists with the performance of the algorithms we hope to gain two way insight: better understanding of the pathologist’s implicit knowledge, and better understanding of the architecture of the learning networks in comparison.
22 +
23 +
The study will be carried out as a collaboration between the groups of Dr Mitra (high throughput neurohistology; machine learning), Dr Kepecs (neurobiology of decision making) and Dr Crawford (clinical histopathology). We plan to set up the eeDAP apparatus in accordance with the instructions provided, assuming we are able to secure the appropriate internal funding. Pathologists from Northwell will serve as subjects of the study.
24 +
25 +
We will coordinate with the IIT Madras team which is also setting up an eeDAP study. The study will make use of the computational framework and web UIs created by the IIT Madras team working collaboratively with Dr Mitra on whole moue brain neurohistological data sets as part of the Mouse Brain Architecture Project.
26 +
27 +
-----------------