A survey on deep learning in medical image analysis

Convolutional Neural Networks (CNNs) have been recently employed to solve problems from both the computer vision and medical image analysis fields. Despite their popularity, most approaches are only able to process 2D images while most medical data used in clinical practice consists of 3D volumes. In this work we propose an approach to 3D image segmentation based on a volumetric, fully convolutional, neural network. Our CNN is trained end-to-end on MRI volumes depicting prostate, and learns to predict segmentation for the whole volume at once. We introduce a novel objective function, that we optimise during training, based on Dice coefficient. In this way we can deal with situations where there is a strong imbalance between the number of foreground and background voxels. To cope with the limited number of annotated volumes available for training, we augment the data applying random non-linear transformations and histogram matching. We show in our experimental evaluation that our approach achieves good performances on challenging test data while requiring only a fraction of the processing time needed by other previous methods.

V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation

We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules of cascaded convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN architectures such as the U-Net model with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed Attention U-Net architecture is evaluated on two large CT abdominal datasets for multi-class image segmentation. Experimental results show that AGs consistently improve the prediction performance of U-Net across different datasets and training sizes while preserving computational efficiency. The code for the proposed architecture is publicly available.

/pdf/attention-u-net-learning-where-to-look-for-the-pancreas-m4jojwwu76.pdf

Attention U-Net: Learning Where to Look for the Pancreas

Artificial intelligence (AI) aims to mimic human cognitive functions. It is bringing a paradigm shift to healthcare, powered by increasing availability of healthcare data and rapid progress of analytics techniques. We survey the current status of AI applications in healthcare and discuss its future. AI can be applied to various types of healthcare data (structured and unstructured). Popular AI techniques include machine learning methods for structured data, such as the classical support vector machine and neural network, and the modern deep learning, as well as natural language processing for unstructured data. Major disease areas that use AI tools include cancer, neurology and cardiology. We then review in more details the AI applications in stroke, in the three major areas of early detection and diagnosis, treatment, as well as outcome prediction and prognosis evaluation. We conclude with discussion about pioneer AI systems, such as IBM Watson, and hurdles for real-life deployment of AI.

https://svn.bmj.com/content/svnbmj/2/4/230.full.pdf

Artificial intelligence in healthcare: past, present and future

https://spiral.imperial.ac.uk/bitstream/10044/1/53839/13/1-s2.0-S1361841516301839-main.pdf

Efficient Multi-Scale 3D CNN with Fully Connected CRF for Accurate Brain Lesion Segmentation

Background: Medical emergency teams (MET) are globally recognized strategies to improve outcomes of deteriorating patients. This study aimed to compare characteristics of hypotension and nonhypotension MET calls by characterizing triggers, timings, risk factors, management strategies, and outcomes of postoperative orthopaedic MET calls and provide information in identifying patients who require further perioperative assessment to facilitate better patient care. Methods: A retrospective, single-center review of postoperative orthopaedic MET calls between October 2017 and April 2019 was performed. MET calls were categorized into three groups depending on MET triggers and analyzed by preoperative, intraoperative, and postoperative parameters. Results: Comparing hypotension-only MET events (Group 1) with other triggers (Group 3), Group 1 showed patients who were younger (71.9 yr vs. 75.8 yr, P=0.012), had elective procedures (52.0% vs. 37.1%, P=0.032), and had non-neck of femoral fracture (71.0% vs. 53.0%, P=0.007). They had fewer comorbidities (Charlson Comorbidity Index: 2.7 vs. 3.6, P=0.011), fewer imaging investigations (12.0% vs. 50.8%, P<0.05), larger volume fluid boluses (0.36L vs. 0.09L, P<0.05), and fewer medication changes (27.0% vs. 43.2%, P=0.013) or additions (18.0% vs. 56.8%, P<0.05). Conclusions: Older age, emergency surgery, higher Charlson Comorbidity Index, preoperative delirium or dementia, and non-neck of femoral fracture orthopaedic patients were at higher risk of nonhypotension METs. In postoperative orthopaedic patients, hypotension-only METs were managed with an average crystalloid fluid bolus of 0.36L. Nonhypotension METs were more likely to be discharged to high level care nursing homes (HLCNH) and less likely to have an ICU admission. These findings have clinical implications for better patient identification and resource allocation of hospital systems. Level of Evidence: Level III.

Joanna P. Simpson

Papers

Efficient Multi-Scale 3D CNN with Fully Connected CRF for Accurate Brain Lesion Segmentation

Nature and characteristics of orthopaedic medical emergency team (MET) events: a retrospective study of 263 MET calls