This Title All WIREs
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 7.250

A 2021 update on cancer image analytics with deep learning

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract Deep learning (DL)‐based interpretation of medical images has reached a critical juncture of expanding outside research projects into translational ones, and is ready to make its way to the clinics. Advances over the last decade in data availability, DL techniques, as well as computing capabilities have accelerated this journey. Through this journey, today we have a better understanding of the challenges to and pitfalls of wider adoption of DL into clinical care, which, according to us, should and will drive the advances in this field in the next few years. The most important among these challenges are the lack of an appropriately digitized environment within healthcare institutions, the lack of adequate open and representative datasets on which DL algorithms can be trained and tested, and the lack of robustness of widely used DL training algorithms to certain pervasive pathological characteristics of medical images and repositories. In this review, we provide an overview of the role of imaging in oncology, the different techniques that are shaping the way DL algorithms are being made ready for clinical use, and also the problems that DL techniques still need to address before DL can find a home in clinics. Finally, we also provide a summary of how DL can potentially drive the adoption of digital pathology, vendor neutral archives, and picture archival and communication systems. We caution that the respective researchers may find the coverage of their own fields to be at a high‐level. This is so by design as this format is meant to only introduce those looking in from outside of deep learning and medical research, respectively, to gain an appreciation for the main concerns and limitations of these two fields instead of telling them something new about their own. This article is categorized under: Technologies > Artificial Intelligence Algorithmic Development > Biological Data Mining
Role of imaging in the clinical journey of a breast cancer patient. A cancer patients' journey is cyclical with individual patients progressing at different rates. Images (pathology and radiology) are acquired at different time points in this journey. The potential applications for deep learning (DL) on individual images are mentioned in solid boxes while applications requiring multitimepoint data are mentioned in empty boxes
[ Normal View | Magnified View ]
An overview of calibrated deep learning model for cancer medical image analysis
[ Normal View | Magnified View ]
An overview of traditional deep learning model for medical image analysis
[ Normal View | Magnified View ]
Common artifacts in CT images. (a) Streak artifact, (b) Ring artifact, and (c) Motion artifact
[ Normal View | Magnified View ]
Common artifacts in histology images: (a) Tissue holes, (b) Tissue fold artifacts, (c) Stain deposits, (d) Knife cutting artifacts, (e) Foreign materials in slide, and (f) Slide artifacts
[ Normal View | Magnified View ]
Schematic diagram illustrating collaborative federated learning framework
[ Normal View | Magnified View ]
Deep multiple instance learning framework for weakly supervised histology image classification
[ Normal View | Magnified View ]
Summary of different categorizations in a machine learning task with respect to calcified nodule. (a) Classification, (b) Object detection, (c) Semantic segmentation, and (d) Instance segmentation of a CT image. Source: TCIA [The Cancer Imaging Archive]
[ Normal View | Magnified View ]
Illustration showing how deep learning has evolved to outperform human capability in image recognition competitions. The results are collected from the ILSVRC, ImageNet challenge (Deng et al., 2009). Top 5 error is a metric used to benchmark the performance of an image recognition model. Here the model is assumed to identify the correct class as one of its top 5 predictions. It is also interesting to note that the large number of model parameters does not necessarily correlate with better performance
[ Normal View | Magnified View ]
Illustration showing a generic experimental setup comprising of two sub‐cohorts (cases and controls) to be predicted (*or a single sub‐cohort). Depending on the desired output type, the cases and controls could have different labels as illustrated in the adjoining table (*or have a continuous label, such as time to event, for deep neural regression)
[ Normal View | Magnified View ]
An example of the temporal flow (left to right) of data collected and decisions made in cancer management such that any combination of the information from the past can be input into a model to make better predictions and decisions about the future. Follow‐up treatments cycle back to the beginning of the flow
[ Normal View | Magnified View ]

Browse by Topic

Algorithmic Development > Biological Data Mining
Technologies > Artificial Intelligence

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts