Dataset:
The grading process consists of finding and classifying cancer tissue into so-called Gleason patterns (3, 4, or 5) based on the architectural growth patterns of the tumour. After the biopsy is assigned a Gleason score, it is converted into an ISUP grade on a 1-5 scale.
The Gleason grading system is the most important prognostic marker for PCa, and the ISUP grade has a crucial role when deciding how a patient should be treated.
There is both a risk of missing cancers and a large risk of over-grading resulting in unnecessary treatment. However, the system suffers from significant inter-observer variability between pathologists, limiting its usefulness for individual patients. This variability in ratings could lead to unnecessary treatment, or worse, missing a severe diagnosis.
Example of the Grading process for a tissue sample
We can improve the accuracy of the process by automating and standardising the recognition of the Gleason Pattern.
- RGB to GrayScale
- Thresholding to convert to Binary Image
- Crop Image - using Polynomial Boundary Representation Principle (to reduce computation and time)
- Closing to enhance Gleason pattern
- Using Local processing to identify Euler's Number of patches
- Create regions using Split And Merge & Euler’s Number
- Identify Majority and Minority Regions
- Output Segmented Tissue Sample and Maj & Min Euler’s Score
Prostate Cancer Grade Detection Notebook
The code generates a segmented tissue sample with an Euler’s score that can be mapped to a Gleason score and ISUP Grade. Illustrated below is a sample output of the above image processing pipeline.
Looking at the Scaled Euler's Score & Area of each Cluster/Segment:
Cluster 1: -5.711280214861236
Cluster 2: -27.291666666666668
Cluster 3: 12.112813370473537
Cluster 4: -13.592105263157896Area of Region 1: 13962500
Area of Region 2: 1500000
Area of Region 3: 4487500
Area of Region 4: 4750000The Pathologist can Identify the 2 distinct most prominent Gleason Patterns and map them to the corresponding Gleason Score & ISUP Grade.
Here we Identified Region 1 - as the Majority Region & Region 2 - as the Minority Region
We can also see that the Euler’s Score estimated for each region is correct as it follows the progression - The lower the Euler’s Number the higher the Gleason Score of the correct Gleason Score i.e. -5.711 is mapped to Score 3 & -27.292 is mapped to Score 4
Therefore the ISUP Grade can be estimated accurately - 3+4 gives Grade 2
- Using the proposed system we were able to Enhance the Tissue Sample and Identify the Gleason Pattern. This included performing operations like - RGB to Grayscale conversion, Boundary Detection, Thresholding, Closing, and so on.
- We used Local Processing to implement Euler’s Number in order to establish a consistent measure to map to the Gleason Scoring System.
- We used the K-Means algorithm for Segmentation, to Identify regions in the tissue sample. We used a custom parameter - the Euler’s Number, instead of the pixel values for Segmentation. We evaluated different values of K to get the most accurate - neither too generalized nor too many regions. We identified K=5 as the best value, (4+1 to account for the background).
- On Further Analysis of the 4 region, the Majority and Minority Gleason Patterns were Identified.
- Then using a Scoring Index these Patterns were mapped to correct Gleason Score.
- The ISUP Grade for the Tissue Sample was Identified.
- Reading in a high resolution (25kX25k pixels) tiff file.
- Reducing Time and Computationally Complexity without Down Sampling
- Converting to Binary Image using the correct Thresholding Approach
- Enhancing Gleason Pattern (Evaluating the performance of different Morphological Operations)
- Calculating Euler's Score for each Region
- Implementing Segmentation using custom value - Euler's Number not the pixel values
- Identifying the correct K value for Segmentation
- Create a dataset using these segmented tissue samples with Gleason Score for each segment
- Build an ML model on that dataset for Prostate Cancer Grade detection