First version

ogemarques · Jul 28, 2021 · 20b683c · 20b683c
1 parent 9d64712
commit 20b683c
Show file tree

Hide file tree

Showing 31 changed files with 64 additions and 1 deletion.
diff --git a/.DS_Store b/.DS_Store
diff --git a/README.md b/README.md
@@ -1 +1,33 @@
-# xai-matlab
+# Explainable AI for Medical Images
+This repository shows an example of how to use MATLAB to produce post-hoc explanations (using [Grad-CAM](https://www.mathworks.com/help/deeplearning/ref/gradcam.html) and [image LIME](https://www.mathworks.com/help/deeplearning/ref/imagelime.html)) for a medical image classification task. 
+
+Both methods (`gradCAM` and `imageLIME`) are available as part of the MATLAB Deep Learning toolbox and require only a single line of code to be applied to results of predictions made by a deep neural network (plus a few lines of code to display the results as a colormap overlaid on the actual images).
+
+| ![](figures/pa_lateral_gradCAM.png) |
+|:--:|
+| Example of `gradCAM` results.|
+
+| ![](figures/pa_lateral_lime.png) |
+|:--:|
+| Example of `imageLIME` results.|
+## Experiment objective
+Given a chest x-ray (CXR), our solution should classify it into Posteroanterior (PA) or Lateral (L) view. 
+### Dataset
+A small subset of the [PadChest](https://bimcv.cipf.es/bimcv-projects/padchest/) dataset<sup>[1](#myfootnote1)</sup>.
+## Requirements
+- [X]  [MATLAB 2020a](https://www.mathworks.com/products/matlab.html) or later
+- [X]  [Deep Learning Toolbox](https://www.mathworks.com/products/deep-learning.html)
+- [X]  [Deep Learning Toolbox™ Model for SqueezeNet Network support package](https://www.mathworks.com/help/deeplearning/ref/squeezenet.html) 
+- [ ]  [Parallel Computing Toolbox](https://www.mathworks.com/products/parallel-computing.html) (only required for training using a GPU)
+## Suggested steps
+1. Download or clone the repository.
+2. Open MATLAB.
+3. Edit the contents of the `dataFolder` variable in the `xai_medical.mlx` file to reflect the path to your selected dataset. 
+4. Run the `xai_medical.mlx` script and inspect results.
+## Additional remarks
+
+- You are encouraged to expand and adapt the example to your needs.
+- The choice of pretrained network and hyperparameters (learning rate, mini-batch size, number of epochs, etc.) is merely illustrative.  
+- You are encouraged to (use Experiment Manager app to) tweak those choices and find a better solution.
+## Notes
+<a name="myfootnote1">[1]</a> This example uses a small subset of images to make it easier to get started without having to worry about large downloads and long training times.   
diff --git a/figures/.DS_Store b/figures/.DS_Store
diff --git a/figures/pa_lateral_gradCAM.png b/figures/pa_lateral_gradCAM.png
diff --git a/figures/pa_lateral_lime.png b/figures/pa_lateral_lime.png
diff --git a/sample_rgb/.DS_Store b/sample_rgb/.DS_Store
diff --git a/sample_rgb/lateral/.DS_Store b/sample_rgb/lateral/.DS_Store
diff --git a/sample_rgb/lateral/126397712012687540784611673197697059691_fp0cu9_imrgb.png b/sample_rgb/lateral/126397712012687540784611673197697059691_fp0cu9_imrgb.png
diff --git a/..._rgb/lateral/216840111366964012373310883942009084123158919_00-069-064_imrgb.png b/..._rgb/lateral/216840111366964012373310883942009084123158919_00-069-064_imrgb.png
diff --git a/..._rgb/lateral/216840111366964012558082906712010004133151165_00-119-134_imrgb.png b/..._rgb/lateral/216840111366964012558082906712010004133151165_00-119-134_imrgb.png
diff --git a/..._rgb/lateral/216840111366964013340662495472012111130929144_01-108-133_imrgb.png b/..._rgb/lateral/216840111366964013340662495472012111130929144_01-108-133_imrgb.png
diff --git a/..._rgb/lateral/216840111366964013451228379692012257110540618_02-006-005_imrgb.png b/..._rgb/lateral/216840111366964013451228379692012257110540618_02-006-005_imrgb.png
diff --git a/..._rgb/lateral/216840111366964014008416513202014153161516671_01-196-149_imrgb.png b/..._rgb/lateral/216840111366964014008416513202014153161516671_01-196-149_imrgb.png
diff --git a/sample_rgb/lateral/317759408657031098460852906419716899073_2rq56f_imrgb.png b/sample_rgb/lateral/317759408657031098460852906419716899073_2rq56f_imrgb.png
diff --git a/sample_rgb/myGrayToRGB.m b/sample_rgb/myGrayToRGB.m
@@ -0,0 +1,31 @@
+function results = myimfcn(im)
+%Image Processing Function
+%
+% IM      - Input image.
+% RESULTS - A scalar structure with the processing results.
+%
+
+%--------------------------------------------------------------------------
+% Auto-generated by imageBatchProcessor App. 
+%
+% When used by the App, this function will be called for every input image
+% file automatically. IM contains the input image as a matrix. RESULTS is a
+% scalar structure containing the results of this processing function.
+%
+%--------------------------------------------------------------------------
+
+
+
+% Replace the sample below with your code----------------------------------
+
+if(size(im,3)==1)
+    % Convert grayscale to RGB
+    imrgb(:,:,1) = im;
+    imrgb(:,:,2) = im;
+    imrgb(:,:,3) = im;
+else
+    imrgb = im;
+end
+
+results.imrgb = imrgb;
+%--------------------------------------------------------------------------