Deep Learning for Magnification Independent Breast Cancer - - PowerPoint PPT Presentation

▶
deep learning for magnification independent breast cancer
SMART_READER_LITE
LIVE PREVIEW

Deep Learning for Magnification Independent Breast Cancer - - PowerPoint PPT Presentation

Nov 07, 2022 361 likes •597 views

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification Neslihan Bayramoglu 1 , Juho Kannala 2 , Janne Heikkil 1 1 Center for Machine Vision and Signal Analysis, University of Oulu, Finland. 2 Department of

slide-1
SLIDE 1

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Neslihan Bayramoglu1, Juho Kannala2, Janne Heikkilä1

1Center for Machine Vision and Signal Analysis,

University of Oulu, Finland.

2Department of Computer Science

Aalto University, Finland

slide-2
SLIDE 2

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

  • Most common cancer among women
  • For a definitive diagnosis
  • Biopsy
  • Microscopic analysis

Breast Cancer

Image: http://www.rafautama.com

slide-3
SLIDE 3

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Image: http://amida13.isi.uu.nl

Sample Preparation

slide-4
SLIDE 4

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Image: www.biomatrix-intl.com

  • Pathologists:
  • pan
  • focus
  • zoom
  • scan can entire image at high magnifications.
  • Timely, costly, and subjective process!

Visual Image Analysis

slide-5
SLIDE 5

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Automated Image Analysis Methods

Computer vision and machine learning methods could automate some of the tasks in the diagnostic pathology workflow

  • Reduce observer variability
  • Increase objectivity
  • Fast and precise quantification
  • Enhance the healthcare quality
slide-6
SLIDE 6

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Challenges

Appearance variability of hematoxylin and eosin stained sections

  • variability among people
  • differences in protocols between labs
  • specimen orientation
  • human skills in tissue preparation
  • microscopy maintenance
  • color variation due to differences in staining procedures
slide-7
SLIDE 7

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Magnification

A malignant breast tumor acquired from a single slide seen in different magnification factors: 40×, 100×, 200×, and 400×

slide-8
SLIDE 8

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Most of the previous studies:

  • Utilize same magnification level

Magnification

Classifier Train data Fixed magnification Model Test image

same

slide-9
SLIDE 9

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Most of the previous studies:

  • Utilize same magnification level

Others:

  • Utilize multiple magnifications
  • Different classifier for each magnification level

Classifier 1

Train data magnification I

Model 1

Test image

s a m e Classifier 2

Train data magnification II

Model 2

Test image

s a m e

Magnification

slide-10
SLIDE 10

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Most of the previous studies:

  • Utilize same magnification level

Others:

  • Utilize multiple magnifications
  • Different classifier for each magnification level

Practical limitations

  • Multiple training stages
  • Test time: magnification factor should be known
  • Difficult to adapt for test images acquired at new magnification

levels.

Magnification

slide-11
SLIDE 11

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Proposal

  • 1. Classify images independent of their magnifications.
  • 2. Multi-task classification

Simultaneous recognition of magnification level and tumor class.

Proposal

slide-12
SLIDE 12

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Dataset

Magnification Benign Malignant Total 40X 652 1370 1995 100X 644 1437 2081 200X 623 1390 2013 400X 588 1232 1820 Total of Images 2480 5429 7909

http://web.inf.ufpr.br/vri/breast-cancer-database

Breast Cancer Histopathological Database (BreakHis)

slide-13
SLIDE 13

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Overview

slide-14
SLIDE 14

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Data augmentation

Augmentation

  • Rotations
  • 90°, 180°, and 270°.
  • Flip
slide-15
SLIDE 15

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Magnification independent classification (single-task CNN)

slide-16
SLIDE 16

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Multi-task classification

slide-17
SLIDE 17

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Multi-task classification

Multi-task loss Each output layer computes a discrete probability distribution by a softmax over the outputs of a fully connected layer.

C = w

bening/malignant

L

  • s

s

bening/malignant

+ w

magnification L

  • s

s

magnification

  • Different weights might improve the results.
  • Difficult to determine theoretically.
  • Needs to be estimated empirically.
slide-18
SLIDE 18

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Experiments and Results

Performance Evaluation NP =Number of images of patient P Nrec= Number of correctly classified images Performance Evaluation

slide-19
SLIDE 19

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Experiments and Results

Recognition Rate (based on patient score) (%) Method/ Magnification 40x 100x 200x 400x Average CLBP 77.4±3.8 76.4±4.5 70.2±3.6 72.8±4.9 74.2 GLCM 74.7±1.0 78.6±2.6 83.4±3.3 81.7±3.3 79.6 LBP 75.6±2.4 73.2±3.5 72.9±2.3 73.1±5.7 73.7 LPQ 73.8±5.0 72.8±5.0 74.3±6.3 73.7±5.7 73.65 ORB 74.4±1.7 69.4±0.4 69.6±3.0 67.6±1.2 70.25 PFTAS 83.8±2.0 82.1±4.9 85.1±3.1 82.3±3.8 83.33 Multi-task CNN 81.87±3.06 83.39±5.17 82.56±3.49 80.69±4.23 82.13 Single-CNN 83.08±2.08 83.17±3.51 84.63±2.72 82.10±4.42 83.25

Hand crafted Features

QDA SVM

slide-20
SLIDE 20

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

Experiments and Results

Recognition Rate (based on patient score) (%) Method/ Magnification 40x 100x 200x 400x Average CLBP 77.4±3.8 76.4±4.5 70.2±3.6 72.8±4.9 74.2 GLCM 74.7±1.0 78.6±2.6 83.4±3.3 81.7±3.3 79.6 LBP 75.6±2.4 73.2±3.5 72.9±2.3 73.1±5.7 73.7 LPQ 73.8±5.0 72.8±5.0 74.3±6.3 73.7±5.7 73.65 ORB 74.4±1.7 69.4±0.4 69.6±3.0 67.6±1.2 70.25 PFTAS 83.8±2.0 82.1±4.9 85.1±3.1 82.3±3.8 83.33 Multi-task CNN 81.87±3.06 83.39±5.17 82.56±3.49 80.69±4.23 82.13 Single-CNN 83.08±2.08 83.17±3.51 84.63±2.72 82.10±4.42 83.25

Hand crafted Features Train 40x Train 100x Train 200x Train 400x Single Train

slide-21
SLIDE 21

Deep Learning for Magnification Independent Breast Cancer Histopathology Image Classification

  • Independent from microscopy magnification and faster than previous

methods.

  • Models are scalable.
  • Multi-task CNN architecture to predict both the image magnification

level and its benign/malignancy property simultaneously.

  • Combine image data from many more resolution levels than four

discrete magnification levels.

  • Magnification level prediction could be formulated as a regression

problem

  • Multi-task prediction requires essentially no additional computation
  • ver single-task prediction

Summary