A timeline of Deep Learning in Digital Pathology

Research in the course of a PhD thesis, by Adrien Foucart.

1. Who was first?

In 2017, Geert Litjens and his colleagues from Radboud University published a comprehensive survey of deep learning methods in medical image analysis [1]. It's possible that they missed some early papers that could qualify as "Deep Learning": as I've written before, the boundaries of what is or isn't Deep Learning are unclear, and articles written before 2012 are unlikely to use that terminology. If such articles exist, however, they have been forgotten in the large pile of never-cited research that fails to be picked-up by Google Scholar, Scopus, or other large research databases. Litjens' survey therefore remains the reference on the matter.

So who was first to use "Deep Learning in Digital Pathology"? The turning point seems to come from the MICCAI 2013 conference in Nagoya, with two pioneering articles: Angel Cruz-Roa's automated skin cancer detection [2], and Dan Cireşan's mitosis detection in breast cancer [3].

Cruz-Roa uses H&E stained images from skin tissue, to try to automatically determine if there is a malignant cancer in the sample. As illustrated in the figure below, the difference between cancer and non-cancer is based on morphological criteria which are very difficult to define.

Example of BCC histopathology
Example of H&E images from cancer and non-cancer skin tissue, taken from [2]

Their algorithm produces both a classification (cancer or not) at the image level, and what they call a "digital staining", which is basically a probability map of where the cancer regions are (see figure below). This is a very important feature for machine learning methods in biomedical imaging, related to the concept of interpretability. A machine learning algorithm which only produces a "diagnosis", but is unable to "explain" how this diagnosis came to be, cannot be trusted. I will most certainly come back to that idea later: the "reasoning" that machine learning algorithms (deep or not) make are sometimes more a reflexion of biases and artefacts in the data that was used to train it than of an understanding of the pathology. Having an output which includes an explanation of the diagnosis is therefore essential to control whether any "weird stuff" is happening.

Output of Cruz-Roa's algorithm
Output of Cruz-Roa's algorithm, taken from [2]

Cireşan also uses H&E images, taken from breast cancer tissue. I'll let him introduce the problem:

Mitosis detection is very hard.

Cireşan, Giusti, Gambardella & Schmidhuber [3]

Mitosis - the process of cell multiplication - is a relatively rare event, meaning that, in the images which are available to train the algorithm, only a very small fraction will be part of a nuclei, and an even smaller fraction will be part of a nuclei going through mitosis. In addition to being rare, the appearance of the cell nucleus will be very different depending of which stage of the mitosis the cell is currently experiencing. To get an idea of how difficult the task is, we can just look at these examples from the article (click on the image for a larger version):

Mitosis and non-mitosis examples
Examples of mitosic and non-mitosic cells, with the "mitosis probability" given by Cireşan's algorithm and the "true classification", taken from [3]

Cireşan's results were far from perfect, but they were impressive enough to be a milestone in the domain: as a winner of the "ICPR 2012 mitosis detection competition", it got a lot of attention... despite the many methodogical issues with the competition itself, which is the topic for another post.

2. Invasion of the Deep Learners

By showing that Deep Learning was a way to get good results on digital pathology tasks, Cireşan and Cruz-Roa opened up the floodgates. Litjens lists many different applications in the subsequent years: bacterial colony counting, classification of mitochondria, classification and grading of different types of cancer, detection of metastases... Mostly on H&E images, but also sometimes using immunohistochemistry, Deep Learning invaded the domain.

A few highly influential works that I would like to mention here, and that I will probably write about more later:

  • In 2015, Olaf Ronneberger, Philipp Fischer and Thomas Brox introduced the U-Net network architecture [4], winning challenges in cell segmentation and cell tracking at the ISBI 2015 conference. This particular architecture is now probably the most widely used in biomedical imaging.
  • In 2016, Andrew Janowczyk and Anant Madabhushi published their Deep Learning for digital pathology tutorial [5], a very practical article on how to approach various use cases in digital pathology, with well-annotated datasets that they also made public. For reasearchers in the field, this is probably one of the best available resource to get started.
  • In 2017, Korsuk Sirinukunwattana published the results of the "GlaS" gland segmentation in colon histology images challenge [6]. This challenge was influential in two ways: first, by providing a high-quality dataset of colon histology images with very accurate annotations, and second by demonstrating how much Deep Learning had penetrated the field of digital pathology. Of the 6 methods deemed good enough to be described in the results article, 5 used deep learning approaches.

3. Looking forward

After these pioneering works, the future of the field may seem a little dull. If we have deep neural network that work for most digital pathology tasks... what is there left to do?

Fortunately, finding a good "deep learning network" is only a part of the "digital pathology pipeline". Everything that comes around the network - from the constitution of the datasets to the way the results are evaluated - is often more important to the final result. That is going to be a large part of what I will write about in future posts. Questions surrounding how data from challenges and data from real-world application may differ, questions about the way we evaluate algorithms, about how we declare winners and losers in ways that may not always reflect how useful the algorithms really are. For that, a good starting point will be to take a closer look at the aforementioned MITOS12 challenge from ICPR 2012.

References

  • [] G. Litjens et al. (2017) "A survey on deep learning in medical image analysis", Medical Image Analysis, 42, 60-88, DOI: 10.1016/j.media.2017.07.005
  • [] A. Cruz-Roa, J. Ovalle, A. Madabhushi, F. Osorio (2013) "A Deep Learning Architecture for Image Representation, Visual Interpretability and Automated Basal-Cell Carcinoma Cancer Detection", Proceedings of MICCAI 2013 in Lecture Notes in Computer Science, 8150, 403-410, DOI: 10.1007/978-3-642-40763-5_50
  • [] D. Cireşan, A. Giusti, L. Gambardella, J. Schmidhuber (2013) "Mitosis Detection in Breast Cancer Histology Images with Deep Neural Networks", Proceedings of MICCAI 2013 in Lecture Notes in Computer Science, 8150, 411-418, DOI: 10.1007/978-3-642-40763-5_51
  • [] O. Ronneberger, P. Fischer, T. Brox (2015) "U-Net: Convolutional Networks for Biomedical Image Segmentation", Proceedings of MICCAI 2015 in Lecture Notes in Computer Science, 9351, 234-241, DOI: 10.1007/978-3-319-24574-4_28
  • [] A. Janowczyk, A. Madabhushi (2016) "Deep learning for digital pathology image analysis: a comprehensive tutorial with selected use cases", Journal of Pathology Informatics, 7, 29, DOI: 10.4103/2153-3539.186902
  • [] K. Sirinukunwattana et al (2017) "Gland Segmentation in Colon Histology Images: The GlaS Challenge Contest", Medical Image Analysis, 35, 489-502, DOI: 10.1016/j.media.2016.08.008