Machine Learning Protocols in Automatic Myelin Segmentation - PowerPoint Presentation

1. Machine Learning Protocols in Automatic Myelin SegmentationPredrag Janjić, MSc.Research Scientist,NIH Research Program in Psychiatric Diseases,RCCSIT@MANUpjanjic@manu.edu.mk

2. Myelin pathology – what do we do ?Myelin pathology studies in our case are part of biological psychiatry research focused on pathology of schizophrenia,conducted within structural studies of “Building Schizophrenia Research in Macedonia”project (PI’s Dwork and Rosoklija), run and coordinated by Neuropathology Lab at Psychiatric Institute of Columbia University in New York.We seek changes in myelin expression at histological level, in post-mortem tissue of diagnosed and control subjects. Apart of structural studies of myelin, the same autopsy samples are used for molecular biology studies of differential myelin expression and its regulation.

3. Who we are ?Andrew Dwork, MD, PhDProfessor of Neuropathology,Psychiatric Institute, Columbia Univ.Gorazd Rosoklija, MD, PhDAssociate Professor of Psychiatry,Psychiatric Institute, Columbia Univ.John Smiley, MD, PhDResearch Scientist, Neurochemistry,Nathan S. Kline Institute, New YorkAleksandar Stankov, MD, PhDForensic Pathologist,Institute of Forensic Medicine,Gordana Petruševska, MD, PhDProfessor of Pathology,Pathology Clinic, School of MedicineBlagoja Dolgoski, MScElectron Microscopist,Project Staff, School of MedicineLjupčo Kocarev, PhDProfessor of Computer Science,Faculty of Informatics and Comp. Eng,RCCSIT at MANUKristijan Petrovski, B.ScSoftware Designer,Project staff at RCCSIT@MANU

4. Image segmentation

5. Myelin – what it is and how it looks like ?Myelin – distinctive fatty structure which wraps and electrically insulates the axons. By preventing the rundown of electrical potential needed for the saltatory conduction, it supports efficient conductivity of neural impulses. Myelin accounts for the major part of brains white-matter.Oligodendroglia is family of cells which members perform myelination of the axons in CNS. Myelin is usually studied through expression of several dominant proteins, fats and enzymes (MBP, PLP1, CNPase, specific glycolipids and several others) which builds up the sheath.Myelination starts in the first year of life and is an intensive process until early adolescence. Myelin degradation is implicated in several diseases and medical conditions.

6. Can we see differences in myelin expression between diagnosed groups and controls ?We measure the myelin content using g-ratio, which simply estimates relative contribution of myelin in fiber area of a single fiber.Due to deformations on bending fibers we use values at low percentile of the myelin thickness distance map (over all measurements), e.g. 5th or 10th percentileWhat questions regarding myelination do we ask ?

7. Segmenting myelin in white matter is challengingAreas implicated in psychiatric diseases have very complex cyto- and myeloarchitecture.Fibers are branching in a wide spread of directions and angles.There are no referent planes of cutting the tissue.Bending fibers / bundles cut at small angles produce “distorted” cross-sections to be measured for both, myelin and axon thickness.Tissue fixation adds several types of artifacts deformations.Zikopoulos & Barbas 2013, Front.Human Neurosci

8. Our histological samplesSamples of human prefrontal white-matter , H&E stained, have been visualized with TEM microscope, with 5.000x magnification.For the development of the method we used 3x 30 images, 2048 x 2048 pixels, 8bit gray-scale.Initial segmentation was attempted in Visiomorph® VIS image processing environment.Segmentation was done in three classes, 1=Myelin, 2=Axon and 3=Background.

9. Machine learning in image processingHow to model and extract morphometric and/or other scale invariant features (SIFT) using computational structure ?Perceptrons Multi-Layer Perceptrons (MLP)Deep NetworksWithin image segmentation, ML algorithm / protocol should classify each pixel in one or more classes

10. Supervised vs. Unsupervised ML in Image ProcessingSupervised learning requires fully annotated learning dataUnsupervised learning usually detects features from examplesClustering methods in unsupervised discriminative learning put objects into different classes

11. Perceptrons and MLP’sClass-1Class-2

12. Computational tools for segmentationTrainable Weka (ImageJ)VIS suite

13. Deep Neural Networks in Image segmentation

14. Annotation of Image Sets for MLNeeded to produce “ground truth” for supervised MLMost laborious step, which puts a limit on efficiency and productivity if to be done manually on large volume of images“Gold-standard” for annotation of histologic samples – three independent annotations (done by microscopist, neuroanatomist and neuropathologist) in all images of the training setCORRECTION

15. Sampling strategies for input dataPixel-based segmentation requires local-model of pixel intensity and neighboring relations (e.g. by convolution of a fragment).Due to statistical nature of ML input data / fragments need statistical sampling.User defined features can, and should bias the sampling, if those are critical for discrimination.Although we introduce a fragment, segmentation the tool assigns a class only to the central pixel, all other pixels just serve as a context

16. Specific sampling strategies for our input dataOptimal window size (33x33, 45x45, 65x65 pixel fragments)Higher statistics of fragments along axon-myelin boundaryEqual representation of small and medium & large fibers according to area,Certain minimal representation of “debris” objects“Padding” at the edges for edge fibersThese and several other interventions are critical for learning efficiency and actual accuracy

17. Learning protocol with Convolutional DNN (ConvNet)A set of 30 full original images are pre-segmented using VIS Segmentation Tool, to get the Interim segmentation status, which reaches average pixel accuracy of 62.4% (variable among the classes).2/3 of the set, 20 images are fragmented into 5M – 8M fragments, usually 45x45 pixels, which are introduced in pairs (Interim, Corrected) in fully supervised manner.Convolution window size varies in narrow range (4 to 6 pixels)There are typically more than one epoch (with same data) needed for the ML tool to saturate the values of interconnection weights5 of remaining images are used for verification which runs along the training, as a separate inputs set whose outputs are used for adjustments, but do not count for overall accuracy.Fragments of 5 last images are used for testing when training is completed, i.e. the cDNN tool is being run as a free classifier where it manifests segmentation capability

18. Initial findings Average accuracy of 88% to 91% percent of pixel accuracy wa achieved (up from ~ 62.4% with VIS),Accuracy of myelin pixels was higher compared to axon and background pixels,Parts of axonal substructure, pre-segmented as myelin were difficult to clean completely,A certain volume of “debris” objects were remaining systematically, due to their structured noise nature (they often look like damaged fibers)

19. Extension of the DNN structure with pre-training layer Instead of random initial values (of weights) in first input layer, we introduced an unsupervised pre-training as a single layer perceptron of specific type (Deep Belief Network),This layer is trained with showing only input fragments from annotated images, which helps to achieve more realistic values for interconnection parameters of that layer once it goes back in DNN training protocolsThis improved dramatically detection and removal of structured noise, i.e. spurious objects

20. Present quality of DNN segmentation protocol

21. How do we measure and average the g-ratio(why do removal of spurious objects matter) ?