Domain             : Computer Vision, Machine Learning
Sub-Domain         : Deep Learning, Image Recognition
Techniques         : Deep Convolutional Neural Network, Transfer Learning, ImageNet, Auto ML, NASNetMobile
Application        : Image Recognition, Image Classification, Medical Imaging

1. Detected Cancer using Auto ML model from Google (“NASNetMobile”) with 250000+ (6.5GB) cancer cell images (histopathologic).
2. For training, concatenated global pooling (max, average), dropout and dense layers to the output layer for final output prediction.
3. Attained testing accuracy of 87.99% and loss of 0.30.

GitHub Link      : Histopathologic Cancer Detection(GitHub)
GitLab Link      : Histopathologic Cancer Detection(GitLab)
Portfolio        : Anjana Tiha's Portfolio

Dataset Name     : Histopathologic Cancer Detection
Dataset Link     : Histopathologic Cancer Detection (Kaggle)
                 : PatchCamelyon (PCam) (GitHub)
                 : CAMELYON16 challenge Dataset (Original Dataset)
                 
Original Paper   : Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer  
                   Authors: Babak Ehteshami Bejnordi, Mitko Veta, Paul Johannes van Diest 
                   JAMA (The Journal of the American Medical Association)
                   Ehteshami Bejnordi B, Veta M, Johannes van Diest P, et al. Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer. JAMA. 2017;318(22):2199–2210. doi:10.1001/jama.2017.14585
                   

Dataset Details
Dataset Name            : Histopathologic Cancer Detection
Number of Class         : 2
Number/Size of Images   : Total      : 220,025 (5.72 Gigabyte (GB))
                          Training   : 132,016 (3.43 Gigabyte (GB))
                          Validation : 44,005  (1.14 Gigabyte (GB))
                          Testing    : 44,004  (1.14 Gigabyte (GB))


Model Parameters
Machine Learning Library: Keras
Base Model              : InceptionV3
Optimizers              : Adam
Loss Function           : categorical_crossentropy


Training Parameters
Batch Size              : 32
Number of Epochs        : 2
Training Time           : 1.5 hour (90 min)


---------------------------------------------------------
Output (Prediction/ Recognition / Classification Metrics)
---------------------------------------------------------

Training
--------------------------------
Accuracy                : 93.59%
Loss                    : 0.1720
--------------------------------

Validation
--------------------------------
Accuracy                : 89.99%
Loss                    : 0.2952
--------------------------------

Testing
--------------------------------
Accuracy                : 89.77%
Loss                    : 89.77%
Precision               : 77.68%
Recall                  : 91.68%
Roc-Auc                 : 86.87%
--------------------------------

Languages               : Python
Tools/IDE               : Anaconda
Libraries               : Keras, TensorFlow, Inception, ImageNet

Duration                : November 2018 - Current
Current Version         : v1.0.0.3
Last Update             : 12.24.2018