Why rankings of biomedical image analysis competitions should be interpreted with care

Automated 3D cytoplasm segmentation in soft X-ray tomography

Cells' structure is key to understanding cellular function, diagnostics, and therapy development. Soft X-ray tomography (SXT) is a unique tool to image cellular structure without fixation or labeling at high spatial resolution and throughput. Fast acquisition times increase demand for accelerated image analysis, like segmentation. Currently, segmenting cellular structures is done manually and is a major bottleneck in the SXT data analysis. This paper introduces ACSeg, an automated 3D cytoplasm segmentation model. ACSeg is generated using semi-automated labels and 3D U-Net and is trained on 43 SXT tomograms of immune T cells, rapidly converging to high-accuracy segmentation, therefore reducing time and labor. Furthermore, adding only 6 SXT tomograms of other cell types diversifies the model, showing potential for optimal experimental design. ACSeg successfully segmented unseen tomograms and is published on Biomedisa, enabling high-throughput analysis of cell volume and structure of cytoplasm in diverse cell types.

RTSeg-net: A lightweight network for real-time segmentation of fetal head and pubic symphysis from intrapartum ultrasound images

The segmentation of the fetal head (FH) and pubic symphysis (PS) from intrapartum ultrasound images plays a pivotal role in monitoring labor progression and informing crucial clinical decisions. Achieving real-time segmentation with high accuracy on systems with limited hardware capabilities presents significant challenges. To address these challenges, we propose the real-time segmentation network (RTSeg-Net), a groundbreaking lightweight deep learning model that incorporates innovative distribution shifting convolutional blocks, tokenized multilayer perceptron blocks, and efficient feature fusion blocks. Designed for optimal computational efficiency, RTSeg-Net minimizes resource demand while significantly enhancing segmentation performance. Our comprehensive evaluation on two distinct intrapartum ultrasound image datasets reveals that RTSeg-Net achieves segmentation accuracy on par with more complex state-of-the-art networks, utilizing merely 1.86 M parameters-just 6 % of their hyperparameters-and operating seven times faster, achieving a remarkable rate of 31.13 frames per second on a Jetson Nano, a device known for its limited computing capacity. These achievements underscore RTSeg-Net's potential to provide accurate, real-time segmentation on low-power devices, broadening the scope for its application across various stages of labor. By facilitating real-time, accurate ultrasound image analysis on portable, low-cost devices, RTSeg-Net promises to revolutionize intrapartum monitoring, making sophisticated diagnostic tools accessible to a wider range of healthcare settings.

Use of superpixels for improvement of inter-rater and intra-rater reliability during annotation of medical images

In the context of automatic medical image segmentation based on statistical learning, raters' variability of ground truth segmentations in training datasets is a widely recognized issue. Indeed, the reference information is provided by experts but bias due to their knowledge may affect the quality of the ground truth data, thus hindering creation of robust and reliable datasets employed in segmentation, classification or detection tasks. In such a framework, automatic medical image segmentation would significantly benefit from utilizing some form of presegmentation during training data preparation process, which could lower the impact of experts' knowledge and reduce time-consuming labeling efforts. The present manuscript proposes a superpixels-driven procedure for annotating medical images. Three different superpixeling methods with two different number of superpixels were evaluated on three different medical segmentation tasks and compared with manual annotations. Within the superpixels-based annotation procedure medical experts interactively select superpixels of interest, apply manual corrections, when necessary, and then the accuracy of the annotations, the time needed to prepare them, and the number of manual corrections are assessed. In this study, it is proven that the proposed procedure reduces inter- and intra-rater variability leading to more reliable annotations datasets which, in turn, may be beneficial for the development of more robust classification or segmentation models. In addition, the proposed approach reduces time needed to prepare the annotations.

Defining and benchmarking open problems in single-cell analysis

With the growing number of single-cell analysis tools, benchmarks are increasingly important to guide analysis and method development. However, a lack of standardisation and extensibility in current benchmarks limits their usability, longevity, and relevance to the community. We present Open Problems, a living, extensible, community-guided benchmarking platform including 10 current single-cell tasks that we envision will raise standards for the selection, evaluation, and development of methods in single-cell analysis.

Bioinformatics in urology - molecular characterization of pathophysiology and response to treatment

The application of bioinformatics has revolutionized the practice of medicine in the past 20 years. From early studies that uncovered subtypes of cancer to broad efforts spearheaded by the Cancer Genome Atlas initiative, the use of bioinformatics strategies to analyse high-dimensional data has provided unprecedented insights into the molecular basis of disease. In addition to the identification of disease subtypes - which enables risk stratification - informatics analysis has facilitated the identification of novel risk factors and drivers of disease, biomarkers of progression and treatment response, as well as possibilities for drug repurposing or repositioning; moreover, bioinformatics has guided research towards precision and personalized medicine. Implementation of specific computational approaches such as artificial intelligence, machine learning and molecular subtyping has yet to become widespread in urology clinical practice for reasons of cost, disruption of clinical workflow and need for prospective validation of informatics approaches in independent patient cohorts. Solving these challenges might accelerate routine integration of bioinformatics into clinical settings.

The multimodality cell segmentation challenge: toward universal solutions

Cell segmentation is a critical step for quantitative single-cell analysis in microscopy images. Existing cell segmentation methods are often tailored to specific modalities or require manual interventions to specify hyper-parameters in different experimental settings. Here, we present a multimodality cell segmentation benchmark, comprising more than 1,500 labeled images derived from more than 50 diverse biological experiments. The top participants developed a Transformer-based deep-learning algorithm that not only exceeds existing methods but can also be applied to diverse microscopy images across imaging platforms and tissue types without manual parameter adjustments. This benchmark and the improved algorithm offer promising avenues for more accurate and versatile cell analysis in microscopy imaging.

Semantic segmentation of thermal defects in belt conveyor idlers using thermal image augmentation and U-Net-based convolutional neural networks

The belt conveyor (BC) is the main means of horizontal transportation of bulk materials at mining sites. The sudden fault in BC modules may cause unexpected stops in production lines. With the increasing number of applications of inspection mobile robots in condition monitoring (CM) of industrial infrastructure in hazardous environments, in this article we introduce an image processing pipeline for automatic segmentation of thermal defects in thermal images captured from BC idlers using a mobile robot. This study follows the fact that CM of idler temperature is an important task for preventing sudden breakdowns in BC system networks. We compared the performance of three different types of U-Net-based convolutional neural network architectures for the identification of thermal anomalies using a small number of hand-labeled thermal images. Experiments on the test data set showed that the attention residual U-Net with binary cross entropy as the loss function handled the semantic segmentation problem better than our previous research and other studied U-Net variations.

Understanding metric-related pitfalls in image analysis validation

Validation metrics are key for tracking scientific progress and bridging the current chasm between artificial intelligence research and its translation into practice. However, increasing evidence shows that, particularly in image analysis, metrics are often chosen inadequately. Although taking into account the individual strengths, weaknesses and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multistage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides a reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Although focused on biomedical image analysis, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. The work serves to enhance global comprehension of a key topic in image analysis validation.

Deep learning for automatic organ and tumor segmentation in nanomedicine pharmacokinetics

Rationale: Multimodal imaging provides important pharmacokinetic and dosimetry information during nanomedicine development and optimization. However, accurate quantitation is time-consuming, resource intensive, and requires anatomical expertise. Methods: We present NanoMASK: a 3D U-Net adapted deep learning tool capable of rapid, automatic organ segmentation of multimodal imaging data that can output key clinical dosimetry metrics without manual intervention. This model was trained on 355 manually-contoured PET/CT data volumes of mice injected with a variety of nanomaterials and imaged over 48 hours. Results: NanoMASK produced 3-dimensional contours of the heart, lungs, liver, spleen, kidneys, and tumor with high volumetric accuracy (pan-organ average %DSC of 92.5). Pharmacokinetic metrics including %ID/cc, %ID, and SUVmax achieved correlation coefficients exceeding R = 0.987 and relative mean errors below 0.2%. NanoMASK was applied to novel datasets of lipid nanoparticles and antibody-drug conjugates with a minimal drop in accuracy, illustrating its generalizability to different classes of nanomedicines. Furthermore, 20 additional auto-segmentation models were developed using training data subsets based on image modality, experimental imaging timepoint, and tumor status. These were used to explore the fundamental biases and dependencies of auto-segmentation models built on a 3D U-Net architecture, revealing significant differential impacts on organ segmentation accuracy. Conclusions: NanoMASK is an easy-to-use, adaptable tool for improving accuracy and throughput in imaging-based pharmacokinetic studies of nanomedicine. It has been made publicly available to all readers for automatic segmentation and pharmacokinetic analysis across a diverse array of nanoparticles, expediting agent development.

Breast Multiparametric MRI for Prediction of Neoadjuvant Chemotherapy Response in Breast Cancer: The BMMR2 Challenge

Purpose To describe the design, conduct, and results of the Breast Multiparametric MRI for prediction of neoadjuvant chemotherapy Response (BMMR2) challenge. Materials and Methods The BMMR2 computational challenge opened on May 28, 2021, and closed on December 21, 2021. The goal of the challenge was to identify image-based markers derived from multiparametric breast MRI, including diffusion-weighted imaging (DWI) and dynamic contrast-enhanced (DCE) MRI, along with clinical data for predicting pathologic complete response (pCR) following neoadjuvant treatment. Data included 573 breast MRI studies from 191 women (mean age [±SD], 48.9 years ± 10.56) in the I-SPY 2/American College of Radiology Imaging Network (ACRIN) 6698 trial (ClinicalTrials.gov: NCT01042379). The challenge cohort was split into training (60%) and test (40%) sets, with teams blinded to test set pCR outcomes. Prediction performance was evaluated by area under the receiver operating characteristic curve (AUC) and compared with the benchmark established from the ACRIN 6698 primary analysis. Results Eight teams submitted final predictions. Entries from three teams had point estimators of AUC that were higher than the benchmark performance (AUC, 0.782 [95% CI: 0.670, 0.893], with AUCs of 0.803 [95% CI: 0.702, 0.904], 0.838 [95% CI: 0.748, 0.928], and 0.840 [95% CI: 0.748, 0.932]). A variety of approaches were used, ranging from extraction of individual features to deep learning and artificial intelligence methods, incorporating DCE and DWI alone or in combination. Conclusion The BMMR2 challenge identified several models with high predictive performance, which may further expand the value of multiparametric breast MRI as an early marker of treatment response. Clinical trial registration no. NCT01042379 Keywords: MRI, Breast, Tumor Response Supplemental material is available for this article. © RSNA, 2024.

The NCI Imaging Data Commons as a platform for reproducible research in computational pathology

BACKGROUND AND OBJECTIVES

Reproducibility is a major challenge in developing machine learning (ML)-based solutions in computational pathology (CompPath). The NCI Imaging Data Commons (IDC) provides >120 cancer image collections according to the FAIR principles and is designed to be used with cloud ML services. Here, we explore its potential to facilitate reproducibility in CompPath research.

METHODS

Using the IDC, we implemented two experiments in which a representative ML-based method for classifying lung tumor tissue was trained and/or evaluated on different datasets. To assess reproducibility, the experiments were run multiple times with separate but identically configured instances of common ML services.

RESULTS

The results of different runs of the same experiment were reproducible to a large extent. However, we observed occasional, small variations in AUC values, indicating a practical limit to reproducibility.

CONCLUSIONS

We conclude that the IDC facilitates approaching the reproducibility limit of CompPath research (i) by enabling researchers to reuse exactly the same datasets and (ii) by integrating with cloud ML services so that experiments can be run in identically configured computing environments.

Automatic Head and Neck Tumor segmentation and outcome prediction relying on FDG-PET/CT images: Findings from the second edition of the HECKTOR challenge

By focusing on metabolic and morphological tissue properties respectively, FluoroDeoxyGlucose (FDG)-Positron Emission Tomography (PET) and Computed Tomography (CT) modalities include complementary and synergistic information for cancerous lesion delineation and characterization (e.g. for outcome prediction), in addition to usual clinical variables. This is especially true in Head and Neck Cancer (HNC). The goal of the HEad and neCK TumOR segmentation and outcome prediction (HECKTOR) challenge was to develop and compare modern image analysis methods to best extract and leverage this information automatically. We present here the post-analysis of HECKTOR 2nd edition, at the 24th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2021. The scope of the challenge was substantially expanded compared to the first edition, by providing a larger population (adding patients from a new clinical center) and proposing an additional task to the challengers, namely the prediction of Progression-Free Survival (PFS). To this end, the participants were given access to a training set of 224 cases from 5 different centers, each with a pre-treatment FDG-PET/CT scan and clinical variables. Their methods were subsequently evaluated on a held-out test set of 101 cases from two centers. For the segmentation task (Task 1), the ranking was based on a Borda counting of their ranks according to two metrics: mean Dice Similarity Coefficient (DSC) and median Hausdorff Distance at 95th percentile (HD95). For the PFS prediction task, challengers could use the tumor contours provided by experts (Task 3) or rely on their own (Task 2). The ranking was obtained according to the Concordance index (C-index) calculated on the predicted risk scores. A total of 103 teams registered for the challenge, for a total of 448 submissions and 29 papers. The best method in the segmentation task obtained an average DSC of 0.759, and the best predictions of PFS obtained a C-index of 0.717 (without relying on the provided contours) and 0.698 (using the expert contours). An interesting finding was that best PFS predictions were reached by relying on DL approaches (with or without explicit tumor segmentation, 4 out of the 5 best ranked) compared to standard radiomics methods using handcrafted features extracted from delineated tumors, and by exploiting alternative tumor contours (automated and/or larger volumes encompassing surrounding tissues) rather than relying on the expert contours. This second edition of the challenge confirmed the promising performance of fully automated primary tumor delineation in PET/CT images of HNC patients, although there is still a margin for improvement in some difficult cases. For the first time, the prediction of outcome was also addressed and the best methods reached relatively good performance (C-index above 0.7). Both results constitute another step forward toward large-scale outcome prediction studies in HNC.

Self-supervised pre-training with contrastive and masked autoencoder methods for dealing with small datasets in deep learning for medical imaging

Deep learning in medical imaging has the potential to minimize the risk of diagnostic errors, reduce radiologist workload, and accelerate diagnosis. Training such deep learning models requires large and accurate datasets, with annotations for all training samples. However, in the medical imaging domain, annotated datasets for specific tasks are often small due to the high complexity of annotations, limited access, or the rarity of diseases. To address this challenge, deep learning models can be pre-trained on large image datasets without annotations using methods from the field of self-supervised learning. After pre-training, small annotated datasets are sufficient to fine-tune the models for a specific task. The most popular self-supervised pre-training approaches in medical imaging are based on contrastive learning. However, recent studies in natural image processing indicate a strong potential for masked autoencoder approaches. Our work compares state-of-the-art contrastive learning methods with the recently introduced masked autoencoder approach "SparK" for convolutional neural networks (CNNs) on medical images. Therefore, we pre-train on a large unannotated CT image dataset and fine-tune on several CT classification tasks. Due to the challenge of obtaining sufficient annotated training data in medical imaging, it is of particular interest to evaluate how the self-supervised pre-training methods perform when fine-tuning on small datasets. By experimenting with gradually reducing the training dataset size for fine-tuning, we find that the reduction has different effects depending on the type of pre-training chosen. The SparK pre-training method is more robust to the training dataset size than the contrastive methods. Based on our results, we propose the SparK pre-training for medical imaging tasks with only small annotated datasets.

Deep learning-based image analysis identifies a DAT-negative subpopulation of dopaminergic neurons in the lateral Substantia nigra

Here we present a deep learning-based image analysis platform (DLAP), tailored to autonomously quantify cell numbers, and fluorescence signals within cellular compartments, derived from RNAscope or immunohistochemistry. We utilised DLAP to analyse subtypes of tyrosine hydroxylase (TH)-positive dopaminergic midbrain neurons in mouse and human brain-sections. These neurons modulate complex behaviour, and are differentially affected in Parkinson's and other diseases. DLAP allows the analysis of large cell numbers, and facilitates the identification of small cellular subpopulations. Using DLAP, we identified a small subpopulation of TH-positive neurons (~5%), mainly located in the very lateral Substantia nigra (SN), that was immunofluorescence-negative for the plasmalemmal dopamine transporter (DAT), with ~40% smaller cell bodies. These neurons were negative for aldehyde dehydrogenase 1A1, with a lower co-expression rate for dopamine-D2-autoreceptors, but a ~7-fold higher likelihood of calbindin-d28k co-expression (~70%). These results have important implications, as DAT is crucial for dopamine signalling, and is commonly used as a marker for dopaminergic SN neurons.

Multiparametric Oncologic Hybrid Imaging: Machine Learning Challenges and Opportunities

BACKGROUND

Machine learning (ML) is considered an important technology for future data analysis in health care.

METHODS

The inherently technology-driven fields of diagnostic radiology and nuclear medicine will both benefit from ML in terms of image acquisition and reconstruction. Within the next few years, this will lead to accelerated image acquisition, improved image quality, a reduction of motion artifacts and - for PET imaging - reduced radiation exposure and new approaches for attenuation correction. Furthermore, ML has the potential to support decision making by a combined analysis of data derived from different modalities, especially in oncology. In this context, we see great potential for ML in multiparametric hybrid imaging and the development of imaging biomarkers.

RESULTS AND CONCLUSION

In this review, we will describe the basics of ML, present approaches in hybrid imaging of MRI, CT, and PET, and discuss the specific challenges associated with it and the steps ahead to make ML a diagnostic and clinical tool in the future.

KEY POINTS

· ML provides a viable clinical solution for the reconstruction, processing, and analysis of hybrid imaging obtained from MRI, CT, and PET..

Natural variability in bee brain size and symmetry revealed by micro-CT imaging and deep learning

Analysing large numbers of brain samples can reveal minor, but statistically and biologically relevant variations in brain morphology that provide critical insights into animal behaviour, ecology and evolution. So far, however, such analyses have required extensive manual effort, which considerably limits the scope for comparative research. Here we used micro-CT imaging and deep learning to perform automated analyses of 3D image data from 187 honey bee and bumblebee brains. We revealed strong inter-individual variations in total brain size that are consistent across colonies and species, and may underpin behavioural variability central to complex social organisations. In addition, the bumblebee dataset showed a significant level of lateralization in optic and antennal lobes, providing a potential explanation for reported variations in visual and olfactory learning. Our fast, robust and user-friendly approach holds considerable promises for carrying out large-scale quantitative neuroanatomical comparisons across a wider range of animals. Ultimately, this will help address fundamental unresolved questions related to the evolution of animal brains and cognition.

Comparing individual and group-level simulated neurophysiological brain connectivity using the Jansen and Rit neural mass model

Computational models are often used to assess how functional connectivity (FC) patterns emerge from neuronal population dynamics and anatomical brain connections. It remains unclear whether the commonly used group-averaged data can predict individual FC patterns. The Jansen and Rit neural mass model was employed, where masses were coupled using individual structural connectivity (SC). Simulated FC was correlated to individual magnetoencephalography-derived empirical FC. FC was estimated using phase-based (phase lag index (PLI), phase locking value (PLV)), and amplitude-based (amplitude envelope correlation (AEC)) metrics to analyze their goodness of fit for individual predictions. Individual FC predictions were compared against group-averaged FC predictions, and we tested whether SC of a different participant could equally well predict participants' FC patterns. The AEC provided a better match between individually simulated and empirical FC than phase-based metrics. Correlations between simulated and empirical FC were higher using individual SC compared to group-averaged SC. Using SC from other participants resulted in similar correlations between simulated and empirical FC compared to using participants' own SC. This work underlines the added value of FC simulations using individual instead of group-averaged SC for this particular computational model and could aid in a better understanding of mechanisms underlying individual functional network trajectories.

AI in medical imaging grand challenges: translation from competition to research benefit and patient care

Artificial intelligence (AI), in one form or another, has been a part of medical imaging for decades. The recent evolution of AI into approaches such as deep learning has dramatically accelerated the application of AI across a wide range of radiologic settings. Despite the promises of AI, developers and users of AI technology must be fully aware of its potential biases and pitfalls, and this knowledge must be incorporated throughout the AI system development pipeline that involves training, validation, and testing. Grand challenges offer an opportunity to advance the development of AI methods for targeted applications and provide a mechanism for both directing and facilitating the development of AI systems. In the process, a grand challenge centralizes (with the challenge organizers) the burden of providing a valid benchmark test set to assess performance and generalizability of participants' models and the collection and curation of image metadata, clinical/demographic information, and the required reference standard. The most relevant grand challenges are those designed to maximize the open-science nature of the competition, with code and trained models deposited for future public access. The ultimate goal of AI grand challenges is to foster the translation of AI systems from competition to research benefit and patient care. Rather than reference the many medical imaging grand challenges that have been organized by groups such as MICCAI, RSNA, AAPM, and grand-challenge.org, this review assesses the role of grand challenges in promoting AI technologies for research advancement and for eventual clinical implementation, including their promises and limitations.

Machine learning for cross-scale microscopy of viruses

Despite advances in virological sciences and antiviral research, viruses continue to emerge, circulate, and threaten public health. We still lack a comprehensive understanding of how cells and individuals remain susceptible to infectious agents. This deficiency is in part due to the complexity of viruses, including the cell states controlling virus-host interactions. Microscopy samples distinct cellular infection stages in a multi-parametric, time-resolved manner at molecular resolution and is increasingly enhanced by machine learning and deep learning. Here we discuss how state-of-the-art artificial intelligence (AI) augments light and electron microscopy and advances virological research of cells. We describe current procedures for image denoising, object segmentation, tracking, classification, and super-resolution and showcase examples of how AI has improved the acquisition and analyses of microscopy data. The power of AI-enhanced microscopy will continue to help unravel virus infection mechanisms, develop antiviral agents, and improve viral vectors.

Automatic generation of subject-specific finite element models of the spine from magnetic resonance images

The generation of subject-specific finite element models of the spine is generally a time-consuming process based on computed tomography (CT) images, where scanning exposes subjects to harmful radiation. In this study, a method is presented for the automatic generation of spine finite element models using images from a single magnetic resonance (MR) sequence. The thoracic and lumbar spine of eight adult volunteers was imaged using a 3D multi-echo-gradient-echo sagittal MR sequence. A deep-learning method was used to generate synthetic CT images from the MR images. A pre-trained deep-learning network was used for the automatic segmentation of vertebrae from the synthetic CT images. Another deep-learning network was trained for the automatic segmentation of intervertebral discs from the MR images. The automatic segmentations were validated against manual segmentations for two subjects, one with scoliosis, and another with a spine implant. A template mesh of the spine was registered to the segmentations in three steps using a Bayesian coherent point drift algorithm. First, rigid registration was applied on the complete spine. Second, non-rigid registration was used for the individual discs and vertebrae. Third, the complete spine was non-rigidly registered to the individually registered discs and vertebrae. Comparison of the automatic and manual segmentations led to dice-scores of 0.93-0.96 for all vertebrae and discs. The lowest dice-score was in the disc at the height of the implant where artifacts led to under-segmentation. The mean distance between the morphed meshes and the segmentations was below 1 mm. In conclusion, the presented method can be used to automatically generate accurate subject-specific spine models.

ROOD-MRI: Benchmarking the robustness of deep learning segmentation models to out-of-distribution and corrupted data in MRI

Deep artificial neural networks (DNNs) have moved to the forefront of medical image analysis due to their success in classification, segmentation, and detection challenges. A principal challenge in large-scale deployment of DNNs in neuroimage analysis is the potential for shifts in signal-to-noise ratio, contrast, resolution, and presence of artifacts from site to site due to variances in scanners and acquisition protocols. DNNs are famously susceptible to these distribution shifts in computer vision. Currently, there are no benchmarking platforms or frameworks to assess the robustness of new and existing models to specific distribution shifts in MRI, and accessible multi-site benchmarking datasets are still scarce or task-specific. To address these limitations, we propose ROOD-MRI: a novel platform for benchmarking the Robustness of DNNs to Out-Of-Distribution (OOD) data, corruptions, and artifacts in MRI. This flexible platform provides modules for generating benchmarking datasets using transforms that model distribution shifts in MRI, implementations of newly derived benchmarking metrics for image segmentation, and examples for using the methodology with new models and tasks. We apply our methodology to hippocampus, ventricle, and white matter hyperintensity segmentation in several large studies, providing the hippocampus dataset as a publicly available benchmark. By evaluating modern DNNs on these datasets, we demonstrate that they are highly susceptible to distribution shifts and corruptions in MRI. We show that while data augmentation strategies can substantially improve robustness to OOD data for anatomical segmentation tasks, modern DNNs using augmentation still lack robustness in more challenging lesion-based segmentation tasks. We finally benchmark U-Nets and vision transformers, finding robustness susceptibility to particular classes of transforms across architectures. The presented open-source platform enables generating new benchmarking datasets and comparing across models to study model design that results in improved robustness to OOD data and corruptions in MRI.

Segmenting functional tissue units across human organs using community-driven development of generalizable machine learning algorithms

The development of a reference atlas of the healthy human body requires automated image segmentation of major anatomical structures across multiple organs based on spatial bioimages generated from various sources with differences in sample preparation. We present the setup and results of the Hacking the Human Body machine learning algorithm development competition hosted by the Human Biomolecular Atlas (HuBMAP) and the Human Protein Atlas (HPA) teams on the Kaggle platform. We create a dataset containing 880 histology images with 12,901 segmented structures, engaging 1175 teams from 78 countries in community-driven, open-science development of machine learning models. Tissue variations in the dataset pose a major challenge to the teams which they overcome by using color normalization techniques and combining vision transformers with convolutional models. The best model will be productized in the HuBMAP portal to process tissue image datasets at scale in support of Human Reference Atlas construction.

Federated benchmarking of medical artificial intelligence with MedPerf

Medical artificial intelligence (AI) has tremendous potential to advance healthcare by supporting and contributing to the evidence-based practice of medicine, personalizing patient treatment, reducing costs, and improving both healthcare provider and patient experience. Unlocking this potential requires systematic, quantitative evaluation of the performance of medical AI models on large-scale, heterogeneous data capturing diverse patient populations. Here, to meet this need, we introduce MedPerf, an open platform for benchmarking AI models in the medical domain. MedPerf focuses on enabling federated evaluation of AI models, by securely distributing them to different facilities, such as healthcare organizations. This process of bringing the model to the data empowers each facility to assess and verify the performance of AI models in an efficient and human-supervised process, while prioritizing privacy. We describe the current challenges healthcare and AI communities face, the need for an open platform, the design philosophy of MedPerf, its current implementation status and real-world deployment, our roadmap and, importantly, the use of MedPerf with multiple international institutions within cloud-based technology and on-premises scenarios. Finally, we welcome new contributions by researchers and organizations to further strengthen MedPerf as an open benchmarking platform.

Investigation and benchmarking of U-Nets on prostate segmentation tasks

In healthcare, a growing number of physicians and support staff are striving to facilitate personalized radiotherapy regimens for patients with prostate cancer. This is because individual patient biology is unique, and employing a single approach for all is inefficient. A crucial step for customizing radiotherapy planning and gaining fundamental information about the disease, is the identification and delineation of targeted structures. However, accurate biomedical image segmentation is time-consuming, requires considerable experience and is prone to observer variability. In the past decade, the use of deep learning models has significantly increased in the field of medical image segmentation. At present, a vast number of anatomical structures can be demarcated on a clinician's level with deep learning models. These models would not only unload work, but they can offer unbiased characterization of the disease. The main architectures used in segmentation are the U-Net and its variants, that exhibit outstanding performances. However, reproducing results or directly comparing methods is often limited by closed source of data and the large heterogeneity among medical images. With this in mind, our intention is to provide a reliable source for assessing deep learning models. As an example, we chose the challenging task of delineating the prostate gland in multi-modal images. First, this paper provides a comprehensive review of current state-of-the-art convolutional neural networks for 3D prostate segmentation. Second, utilizing public and in-house CT and MR datasets of varying properties, we created a framework for an objective comparison of automatic prostate segmentation algorithms. The framework was used for rigorous evaluations of the models, highlighting their strengths and weaknesses.

BigNeuron: a resource to benchmark and predict performance of algorithms for automated tracing of neurons in light microscopy datasets

BigNeuron is an open community bench-testing platform with the goal of setting open standards for accurate and fast automatic neuron tracing. We gathered a diverse set of image volumes across several species that is representative of the data obtained in many neuroscience laboratories interested in neuron tracing. Here, we report generated gold standard manual annotations for a subset of the available imaging datasets and quantified tracing quality for 35 automatic tracing algorithms. The goal of generating such a hand-curated diverse dataset is to advance the development of tracing algorithms and enable generalizable benchmarking. Together with image quality features, we pooled the data in an interactive web application that enables users and developers to perform principal component analysis, t-distributed stochastic neighbor embedding, correlation and clustering, visualization of imaging and tracing data, and benchmarking of automatic tracing algorithms in user-defined data subsets. The image quality metrics explain most of the variance in the data, followed by neuromorphological features related to neuron size. We observed that diverse algorithms can provide complementary information to obtain accurate results and developed a method to iteratively combine methods and generate consensus reconstructions. The consensus trees obtained provide estimates of the neuron structure ground truth that typically outperform single algorithms in noisy datasets. However, specific algorithms may outperform the consensus tree strategy in specific imaging conditions. Finally, to aid users in predicting the most accurate automatic tracing results without manual annotations for comparison, we used support vector machine regression to predict reconstruction quality given an image volume and a set of automatic tracings.

PEg TRAnsfer Workflow recognition challenge report: Do multimodal data improve recognition?

BACKGROUND AND OBJECTIVE

In order to be context-aware, computer-assisted surgical systems require accurate, real-time automatic surgical workflow recognition. In the past several years, surgical video has been the most commonly-used modality for surgical workflow recognition. But with the democratization of robot-assisted surgery, new modalities, such as kinematics, are now accessible. Some previous methods use these new modalities as input for their models, but their added value has rarely been studied. This paper presents the design and results of the "PEg TRAnsfer Workflow recognition" (PETRAW) challenge with the objective of developing surgical workflow recognition methods based on one or more modalities and studying their added value.

METHODS

The PETRAW challenge included a data set of 150 peg transfer sequences performed on a virtual simulator. This data set included videos, kinematic data, semantic segmentation data, and annotations, which described the workflow at three levels of granularity: phase, step, and activity. Five tasks were proposed to the participants: three were related to the recognition at all granularities simultaneously using a single modality, and two addressed the recognition using multiple modalities. The mean application-dependent balanced accuracy (AD-Accuracy) was used as an evaluation metric to take into account class balance and is more clinically relevant than a frame-by-frame score.

RESULTS

Seven teams participated in at least one task with four participating in every task. The best results were obtained by combining video and kinematic data (AD-Accuracy of between 93% and 90% for the four teams that participated in all tasks).

CONCLUSION

The improvement of surgical workflow recognition methods using multiple modalities compared with unimodal methods was significant for all teams. However, the longer execution time required for video/kinematic-based methods(compared to only kinematic-based methods) must be considered. Indeed, one must ask if it is wise to increase computing time by 2000 to 20,000% only to increase accuracy by 3%. The PETRAW data set is publicly available at www.synapse.org/PETRAW to encourage further research in surgical workflow recognition.

Smartphone Integration of Artificial Intelligence for Automated Plagiocephaly Diagnosis

Positional plagiocephaly is a pediatric condition with important cosmetic implications affecting ∼40% of infants under 12 months of age. Early diagnosis and treatment initiation is imperative in achieving satisfactory outcomes; improved diagnostic modalities are needed to support this goal. This study aimed to determine whether a smartphone-based artificial intelligence tool could diagnose positional plagiocephaly.

Methods

A prospective validation study was conducted at a large tertiary care center with two recruitment sites: (1) newborn nursery, (2) pediatric craniofacial surgery clinic. Eligible children were aged 0-12 months with no history of hydrocephalus, intracranial tumors, intracranial hemorrhage, intracranial hardware, or prior craniofacial surgery. Successful artificial intelligence diagnosis required identification of the presence and severity of positional plagiocephaly.

Results

A total of 89 infants were prospectively enrolled from the craniofacial surgery clinic (n = 25, 17 male infants [68%], eight female infants [32%], mean age 8.44 months) and newborn nursery (n = 64, 29 male infants [45%], 25 female infants [39%], mean age 0 months). The model obtained a diagnostic accuracy of 85.39% compared with a standard clinical examination with a disease prevalence of 48%. Sensitivity was 87.50% [95% CI, 75.94-98.42] with a specificity of 83.67% [95% CI, 72.35-94.99]. Precision was 81.40%, while likelihood ratios (positive and negative) were 5.36 and 0.15, respectively. The F1-score was 84.34%.

Conclusions

The smartphone-based artificial intelligence algorithm accurately diagnosed positional plagiocephaly in a clinical environment. This technology may provide value by helping guide specialist consultation and enabling longitudinal quantitative monitoring of cranial shape.

Enhanced Pathology Image Quality with Restore-Generative Adversarial Network

Whole slide imaging is becoming a routine procedure in clinical diagnosis. Advanced image analysis techniques have been developed to assist pathologists in disease diagnosis, staging, subtype classification, and risk stratification. Recently, deep learning algorithms have achieved state-of-the-art performances in various imaging analysis tasks, including tumor region segmentation, nuclei detection, and disease classification. However, widespread clinical use of these algorithms is hampered by their performances often degrading due to image quality issues commonly seen in real-world pathology imaging data such as low resolution, blurring regions, and staining variation. Restore-Generative Adversarial Network (GAN), a deep learning model, was developed to improve the imaging qualities by restoring blurred regions, enhancing low resolution, and normalizing staining colors. The results demonstrate that Restore-GAN can significantly improve image quality, which leads to improved model robustness and performance for existing deep learning algorithms in pathology image analysis. Restore-GAN has the potential to be used to facilitate the applications of deep learning models in digital pathology analyses.

Assessing Interobserver Variability in the Delineation of Structures in Radiation Oncology: A Systematic Review

PURPOSE

The delineation of target volumes and organs at risk is the main source of uncertainty in radiation therapy. Numerous interobserver variability (IOV) studies have been conducted, often with unclear methodology and nonstandardized reporting. We aimed to identify the parameters chosen in conducting delineation IOV studies and assess their performances and limits.

METHODS AND MATERIALS

We conducted a systematic literature review to highlight major points of heterogeneity and missing data in IOV studies published between 2018 and 2021. For the main used metrics, we did in silico analyses to assess their limits in specific clinical situations.

RESULTS

All disease sites were represented in the 66 studies examined. Organs at risk were studied independently of tumor site in 29% of reviewed IOV studies. In 65% of studies, statistical analyses were performed. No gold standard (GS; ie, reference) was defined in 36% of studies. A single expert was considered as the GS in 21% of studies, without testing intraobserver variability. All studies reported both absolute and relative indices, including the Dice similarity coefficient (DSC) in 68% and the Hausdorff distance (HD) in 42%. Limitations were shown in silico for small structures when using the DSC and dependence on irregular shapes when using the HD. Variations in DSC values were large between studies, and their thresholds were inconsistent. Most studies (51%) included 1 to 10 cases. The median number of observers or experts was 7 (range, 2-35). The intraclass correlation coefficient was reported in only 9% of cases. Investigating the feasibility of studying IOV in delineation, a minimum of 8 observers with 3 cases, or 11 observers with 2 cases, was required to demonstrate moderate reproducibility.

CONCLUSIONS

Implementation of future IOV studies would benefit from a more standardized methodology: clear definitions of the gold standard and metrics and a justification of the tradeoffs made in the choice of the number of observers and number of delineated cases should be provided.

Analysis: Flawed Datasets of Monkeypox Skin Images

The self-proclaimed first publicly available dataset of Monkeypox skin images consists of medically irrelevant images extracted from Google and photography repositories through a process denominated web-scrapping. Yet, this did not stop other researchers from employing it to build Machine Learning (ML) solutions aimed at computer-aided diagnosis of Monkeypox and other viral infections presenting skin lesions. Neither did it stop the reviewers or editors from publishing these subsequent works in peer-reviewed journals. Several of these works claimed extraordinary performance in the classification of Monkeypox, Chickenpox and Measles, employing ML and the aforementioned dataset. In this work, we analyse the initiator work that has catalysed the development of several ML solutions, and whose popularity is continuing to grow. Further, we provide a rebuttal experiment that showcases the risks of such methodologies, proving that the ML solutions do not necessarily obtain their performance from the features relevant to the diseases at issue.

Labelling instructions matter in biomedical image analysis

Biomedical image analysis algorithm validation depends on high-quality annotation of reference datasets, for which labelling instructions are key. Despite their importance, their optimization remains largely unexplored. Here we present a systematic study of labelling instructions and their impact on annotation quality in the field. Through comprehensive examination of professional practice and international competitions registered at the Medical Image Computing and Computer Assisted Intervention Society, the largest international society in the biomedical imaging field, we uncovered a discrepancy between annotators’ needs for labelling instructions and their current quality and availability. On the basis of an analysis of 14,040 images annotated by 156 annotators from four professional annotation companies and 708 Amazon Mechanical Turk crowdworkers using instructions with different information density levels, we further found that including exemplary images substantially boosts annotation performance compared with text-only descriptions, while solely extending text descriptions does not. Finally, professional annotators constantly outperform Amazon Mechanical Turk crowdworkers. Our study raises awareness for the need of quality standards in biomedical image analysis labelling instructions.

Beyond rankings: Learning (more) from algorithm validation

Challenges have become the state-of-the-art approach to benchmark image analysis algorithms in a comparative manner. While the validation on identical data sets was a great step forward, results analysis is often restricted to pure ranking tables, leaving relevant questions unanswered. Specifically, little effort has been put into the systematic investigation on what characterizes images in which state-of-the-art algorithms fail. To address this gap in the literature, we (1) present a statistical framework for learning from challenges and (2) instantiate it for the specific task of instrument instance segmentation in laparoscopic videos. Our framework relies on the semantic meta data annotation of images, which serves as foundation for a General Linear Mixed Models (GLMM) analysis. Based on 51,542 meta data annotations performed on 2,728 images, we applied our approach to the results of the Robust Medical Instrument Segmentation Challenge (ROBUST-MIS) challenge 2019 and revealed underexposure, motion and occlusion of instruments as well as the presence of smoke or other objects in the background as major sources of algorithm failure. Our subsequent method development, tailored to the specific remaining issues, yielded a deep learning model with state-of-the-art overall performance and specific strengths in the processing of images in which previous methods tended to fail. Due to the objectivity and generic applicability of our approach, it could become a valuable tool for validation in the field of medical image analysis and beyond.

Artificial intelligence and automation in endoscopy and surgery

Modern endoscopy relies on digital technology, from high-resolution imaging sensors and displays to electronics connecting configurable illumination and actuation systems for robotic articulation. In addition to enabling more effective diagnostic and therapeutic interventions, the digitization of the procedural toolset enables video data capture of the internal human anatomy at unprecedented levels. Interventional video data encapsulate functional and structural information about a patient's anatomy as well as events, activity and action logs about the surgical process. This detailed but difficult-to-interpret record from endoscopic procedures can be linked to preoperative and postoperative records or patient imaging information. Rapid advances in artificial intelligence, especially in supervised deep learning, can utilize data from endoscopic procedures to develop systems for assisting procedures leading to computer-assisted interventions that can enable better navigation during procedures, automation of image interpretation and robotically assisted tool manipulation. In this Perspective, we summarize state-of-the-art artificial intelligence for computer-assisted interventions in gastroenterology and surgery.

Comparative validation of machine learning algorithms for surgical workflow and skill analysis with the HeiChole benchmark

PURPOSE

Surgical workflow and skill analysis are key technologies for the next generation of cognitive surgical assistance systems. These systems could increase the safety of the operation through context-sensitive warnings and semi-autonomous robotic assistance or improve training of surgeons via data-driven feedback. In surgical workflow analysis up to 91% average precision has been reported for phase recognition on an open data single-center video dataset. In this work we investigated the generalizability of phase recognition algorithms in a multicenter setting including more difficult recognition tasks such as surgical action and surgical skill.

METHODS

To achieve this goal, a dataset with 33 laparoscopic cholecystectomy videos from three surgical centers with a total operation time of 22 h was created. Labels included framewise annotation of seven surgical phases with 250 phase transitions, 5514 occurences of four surgical actions, 6980 occurences of 21 surgical instruments from seven instrument categories and 495 skill classifications in five skill dimensions. The dataset was used in the 2019 international Endoscopic Vision challenge, sub-challenge for surgical workflow and skill analysis. Here, 12 research teams trained and submitted their machine learning algorithms for recognition of phase, action, instrument and/or skill assessment.

RESULTS

F1-scores were achieved for phase recognition between 23.9% and 67.7% (n = 9 teams), for instrument presence detection between 38.5% and 63.8% (n = 8 teams), but for action recognition only between 21.8% and 23.3% (n = 5 teams). The average absolute error for skill assessment was 0.78 (n = 1 team).

CONCLUSION

Surgical workflow and skill analysis are promising technologies to support the surgical team, but there is still room for improvement, as shown by our comparison of machine learning algorithms. This novel HeiChole benchmark can be used for comparable evaluation and validation of future work. In future studies, it is of utmost importance to create more open, high-quality datasets in order to allow the development of artificial intelligence and cognitive robotics in surgery.

A 178-clinical-center experiment of integrating AI solutions for lung pathology diagnosis

In 2020, an experiment testing AI solutions for lung X-ray analysis on a multi-hospital network was conducted. The multi-hospital network linked 178 Moscow state healthcare centers, where all chest X-rays from the network were redirected to a research facility, analyzed with AI, and returned to the centers. The experiment was formulated as a public competition with monetary awards for participating industrial and research teams. The task was to perform the binary detection of abnormalities from chest X-rays. For the objective real-life evaluation, no training X-rays were provided to the participants. This paper presents one of the top-performing AI frameworks from this experiment. First, the framework used two EfficientNets, histograms of gradients, Haar feature ensembles, and local binary patterns to recognize whether an input image represents an acceptable lung X-ray sample, meaning the X-ray is not grayscale inverted, is a frontal chest X-ray, and completely captures both lung fields. Second, the framework extracted the region with lung fields and then passed them to a multi-head DenseNet, where the heads recognized the patient's gender, age and the potential presence of abnormalities, and generated the heatmap with the abnormality regions highlighted. During one month of the experiment from 11.23.2020 to 12.25.2020, 17,888 cases have been analyzed by the framework with 11,902 cases having radiological reports with the reference diagnoses that were unequivocally parsed by the experiment organizers. The performance measured in terms of the area under receiving operator curve (AUC) was 0.77. The AUC for individual diseases ranged from 0.55 for herniation to 0.90 for pneumothorax.

Shortcomings and areas for improvement in digital pathology image segmentation challenges

Digital pathology image analysis challenges have been organised regularly since 2010, often with events hosted at major conferences and results published in high-impact journals. These challenges mobilise a lot of energy from organisers, participants, and expert annotators (especially for image segmentation challenges). This study reviews image segmentation challenges in digital pathology and the top-ranked methods, with a particular focus on how reference annotations are generated and how the methods' predictions are evaluated. We found important shortcomings in the handling of inter-expert disagreement and the relevance of the evaluation process chosen. We also noted key problems with the quality control of various challenge elements that can lead to uncertainties in the published results. Our findings show the importance of greatly increasing transparency in the reporting of challenge results, and the need to make publicly available the evaluation codes, test set annotations and participants' predictions. The aim is to properly ensure the reproducibility and interpretation of the results and to increase the potential for exploitation of the substantial work done in these challenges.

The KNee OsteoArthritis Prediction (KNOAP2020) challenge: An image analysis challenge to predict incident symptomatic radiographic knee osteoarthritis from MRI and X-ray images

OBJECTIVES

The KNee OsteoArthritis Prediction (KNOAP2020) challenge was organized to objectively compare methods for the prediction of incident symptomatic radiographic knee osteoarthritis within 78 months on a test set with blinded ground truth.

DESIGN

The challenge participants were free to use any available data sources to train their models. A test set of 423 knees from the Prevention of Knee Osteoarthritis in Overweight Females (PROOF) study consisting of magnetic resonance imaging (MRI) and X-ray image data along with clinical risk factors at baseline was made available to all challenge participants. The ground truth outcomes, i.e., which knees developed incident symptomatic radiographic knee osteoarthritis (according to the combined ACR criteria) within 78 months, were not provided to the participants. To assess the performance of the submitted models, we used the area under the receiver operating characteristic curve (ROCAUC) and balanced accuracy (BACC).

RESULTS

Seven teams submitted 23 entries in total. A majority of the algorithms were trained on data from the Osteoarthritis Initiative. The model with the highest ROCAUC (0.64 (95% confidence interval (CI): 0.57-0.70)) used deep learning to extract information from X-ray images combined with clinical variables. The model with the highest BACC (0.59 (95% CI: 0.52-0.65)) ensembled three different models that used automatically extracted X-ray and MRI features along with clinical variables.

CONCLUSION

The KNOAP2020 challenge established a benchmark for predicting incident symptomatic radiographic knee osteoarthritis. Accurate prediction of incident symptomatic radiographic knee osteoarthritis is a complex and still unsolved problem requiring additional investigation.

CrossMoDA 2021 challenge: Benchmark of cross-modality domain adaptation techniques for vestibular schwannoma and cochlea segmentation

Domain Adaptation (DA) has recently been of strong interest in the medical imaging community. While a large variety of DA techniques have been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality Domain Adaptation (crossMoDA) challenge was organised in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2021). CrossMoDA is the first large and multi-class benchmark for unsupervised cross-modality Domain Adaptation. The goal of the challenge is to segment two key brain structures involved in the follow-up and treatment planning of vestibular schwannoma (VS): the VS and the cochleas. Currently, the diagnosis and surveillance in patients with VS are commonly performed using contrast-enhanced T1 (ceT₁) MR imaging. However, there is growing interest in using non-contrast imaging sequences such as high-resolution T2 (hrT₂) imaging. For this reason, we established an unsupervised cross-modality segmentation benchmark. The training dataset provides annotated ceT₁ scans (N=105) and unpaired non-annotated hrT₂ scans (N=105). The aim was to automatically perform unilateral VS and bilateral cochlea segmentation on hrT₂ scans as provided in the testing set (N=137). This problem is particularly challenging given the large intensity distribution gap across the modalities and the small volume of the structures. A total of 55 teams from 16 countries submitted predictions to the validation leaderboard. Among them, 16 teams from 9 different countries submitted their algorithm for the evaluation phase. The level of performance reached by the top-performing teams is strikingly high (best median Dice score - VS: 88.4%; Cochleas: 85.7%) and close to full supervision (median Dice score - VS: 92.5%; Cochleas: 87.7%). All top-performing methods made use of an image-to-image translation approach to transform the source-domain images into pseudo-target-domain images. A segmentation network was then trained using these generated images and the manual annotations provided for the source image.

Challenges for machine learning in clinical translation of big data imaging studies

Combining deep learning image analysis methods and large-scale imaging datasets offers many opportunities to neuroscience imaging and epidemiology. However, despite these opportunities and the success of deep learning when applied to a range of neuroimaging tasks and domains, significant barriers continue to limit the impact of large-scale datasets and analysis tools. Here, we examine the main challenges and the approaches that have been explored to overcome them. We focus on issues relating to data availability, interpretability, evaluation, and logistical challenges and discuss the problems that still need to be tackled to enable the success of "big data" deep learning approaches beyond research.

Reproducibility of deep learning in digital pathology whole slide image analysis

For a method to be widely adopted in medical research or clinical practice, it needs to be reproducible so that clinicians and regulators can have confidence in its use. Machine learning and deep learning have a particular set of challenges around reproducibility. Small differences in the settings or the data used for training a model can lead to large differences in the outcomes of experiments. In this work, three top-performing algorithms from the Camelyon grand challenges are reproduced using only information presented in the associated papers and the results are then compared to those reported. Seemingly minor details were found to be critical to performance and yet their importance is difficult to appreciate until the actual reproduction is attempted. We observed that authors generally describe the key technical aspects of their models well but fail to maintain the same reporting standards when it comes to data preprocessing which is essential to reproducibility. As an important contribution of the present study and its findings, we introduce a reproducibility checklist that tabulates information that needs to be reported in histopathology ML-based work in order to make it reproducible.

Multicentric exploration of tool annotation in robotic surgery: lessons learned when starting a surgical artificial intelligence project

BACKGROUND

Artificial intelligence (AI) holds tremendous potential to reduce surgical risks and improve surgical assessment. Machine learning, a subfield of AI, can be used to analyze surgical video and imaging data. Manual annotations provide veracity about the desired target features. Yet, methodological annotation explorations are limited to date. Here, we provide an exploratory analysis of the requirements and methods of instrument annotation in a multi-institutional team from two specialized AI centers and compile our lessons learned.

METHODS

We developed a bottom-up approach for team annotation of robotic instruments in robot-assisted partial nephrectomy (RAPN), which was subsequently validated in robot-assisted minimally invasive esophagectomy (RAMIE). Furthermore, instrument annotation methods were evaluated for their use in Machine Learning algorithms. Overall, we evaluated the efficiency and transferability of the proposed team approach and quantified performance metrics (e.g., time per frame required for each annotation modality) between RAPN and RAMIE.

RESULTS

We found a 0.05 Hz image sampling frequency to be adequate for instrument annotation. The bottom-up approach in annotation training and management resulted in accurate annotations and demonstrated efficiency in annotating large datasets. The proposed annotation methodology was transferrable between both RAPN and RAMIE. The average annotation time for RAPN pixel annotation ranged from 4.49 to 12.6 min per image; for vector annotation, we denote 2.92 min per image. Similar annotation times were found for RAMIE. Lastly, we elaborate on common pitfalls encountered throughout the annotation process.

CONCLUSIONS

We propose a successful bottom-up approach for annotator team composition, applicable to any surgical annotation project. Our results set the foundation to start AI projects for instrument detection, segmentation, and pose estimation. Due to the immense annotation burden resulting from spatial instrumental annotation, further analysis into sampling frequency and annotation detail needs to be conducted.

Inflation of test accuracy due to data leakage in deep learning-based classification of OCT images

In the application of deep learning on optical coherence tomography (OCT) data, it is common to train classification networks using 2D images originating from volumetric data. Given the micrometer resolution of OCT systems, consecutive images are often very similar in both visible structures and noise. Thus, an inappropriate data split can result in overlap between the training and testing sets, with a large portion of the literature overlooking this aspect. In this study, the effect of improper dataset splitting on model evaluation is demonstrated for three classification tasks using three OCT open-access datasets extensively used, Kermany's and Srinivasan's ophthalmology datasets, and AIIMS breast tissue dataset. Results show that the classification performance is inflated by 0.07 up to 0.43 in terms of Matthews Correlation Coefficient (accuracy: 5% to 30%) for models tested on datasets with improper splitting, highlighting the considerable effect of dataset handling on model evaluation. This study intends to raise awareness on the importance of dataset splitting given the increased research interest in implementing deep learning on OCT data.

Rapid artificial intelligence solutions in a pandemic-The COVID-19-20 Lung CT Lesion Segmentation Challenge

Artificial intelligence (AI) methods for the automatic detection and quantification of COVID-19 lesions in chest computed tomography (CT) might play an important role in the monitoring and management of the disease. We organized an international challenge and competition for the development and comparison of AI algorithms for this task, which we supported with public data and state-of-the-art benchmark methods. Board Certified Radiologists annotated 295 public images from two sources (A and B) for algorithms training (n=199, source A), validation (n=50, source A) and testing (n=23, source A; n=23, source B). There were 1,096 registered teams of which 225 and 98 completed the validation and testing phases, respectively. The challenge showed that AI models could be rapidly designed by diverse teams with the potential to measure disease or facilitate timely and patient-specific interventions. This paper provides an overview and the major outcomes of the COVID-19 Lung CT Lesion Segmentation Challenge — 2020.

Deep learning-based automated segmentation of resection cavities on postsurgical epilepsy MRI

Accurate segmentation of surgical resection sites is critical for clinical assessments and neuroimaging research applications, including resection extent determination, predictive modeling of surgery outcome, and masking image processing near resection sites. In this study, an automated resection cavity segmentation algorithm is developed for analyzing postoperative MRI of epilepsy patients and deployed in an easy-to-use graphical user interface (GUI) that estimates remnant brain volumes, including postsurgical hippocampal remnant tissue. This retrospective study included postoperative T1-weighted MRI from 62 temporal lobe epilepsy (TLE) patients who underwent resective surgery. The resection site was manually segmented and reviewed by a neuroradiologist (JMS). A majority vote ensemble algorithm was used to segment surgical resections, using 3 U-Net convolutional neural networks trained on axial, coronal, and sagittal slices, respectively. The algorithm was trained using 5-fold cross validation, with data partitioned into training (N = 27) testing (N = 9), and validation (N = 9) sets, and evaluated on a separate held-out test set (N = 17). Algorithm performance was assessed using Dice-Sørensen coefficient (DSC), Hausdorff distance, and volume estimates. Additionally, we deploy a fully-automated, GUI-based pipeline that compares resection segmentations with preoperative imaging and reports estimates of resected brain structures. The cross-validation and held-out test median DSCs were 0.84 ± 0.08 and 0.74 ± 0.22 (median ± interquartile range) respectively, which approach inter-rater reliability between radiologists (0.84-0.86) as reported in the literature. Median 95 % Hausdorff distances were 3.6 mm and 4.0 mm respectively, indicating high segmentation boundary confidence. Automated and manual resection volume estimates were highly correlated for both cross-validation (r = 0.94, p < 0.0001) and held-out test subjects (r = 0.87, p < 0.0001). Automated and manual segmentations overlapped in all 62 subjects, indicating a low false negative rate. In control subjects (N = 40), the classifier segmented no voxels (N = 33), <50 voxels (N = 5), or a small volumes<0.5 cm3 (N = 2), indicating a low false positive rate that can be controlled via thresholding. There was strong agreement between postoperative hippocampal remnant volumes determined using automated and manual resection segmentations (r = 0.90, p < 0.0001, mean absolute error = 6.3 %), indicating that automated resection segmentations can permit quantification of postoperative brain volumes after epilepsy surgery. Applications include quantification of postoperative remnant brain volumes, correction of deformable registration, and localization of removed brain regions for network modeling.

The Medical Segmentation Decathlon

International challenges have become the de facto standard for comparative assessment of image analysis algorithms. Although segmentation is the most widely investigated medical image processing task, the various challenges have been organized to focus only on specific clinical tasks. We organized the Medical Segmentation Decathlon (MSD)-a biomedical image analysis challenge, in which algorithms compete in a multitude of both tasks and modalities to investigate the hypothesis that a method capable of performing well on multiple tasks will generalize well to a previously unseen task and potentially outperform a custom-designed solution. MSD results confirmed this hypothesis, moreover, MSD winner continued generalizing well to a wide range of other clinical problems for the next two years. Three main conclusions can be drawn from this study: (1) state-of-the-art image segmentation algorithms generalize well when retrained on unseen tasks; (2) consistent algorithmic performance across multiple tasks is a strong surrogate of algorithmic generalizability; (3) the training of accurate AI segmentation models is now commoditized to scientists that are not versed in AI model training.

Accuracy of Automated Computer-Aided Diagnosis for Stroke Imaging: A Critical Evaluation of Current Evidence

There is increasing interest in computer applications, using artificial intelligence methodologies, to perform health care tasks previously performed by humans, particularly in medical imaging for diagnosis. In stroke, there are now commercial artificial intelligence software for use with computed tomography or MR imaging to identify acute ischemic brain tissue pathology, arterial obstruction on computed tomography angiography or as hyperattenuated arteries on computed tomography, brain hemorrhage, or size of perfusion defects. A rapid, accurate diagnosis may aid treatment decisions for individual patients and could improve outcome if it leads to effective and safe treatment; or conversely, to disaster if a delayed or incorrect diagnosis results in inappropriate treatment. Despite this potential clinical impact, diagnostic tools including artificial intelligence methods are not subjected to the same clinical evaluation standards as are mandatory for drugs. Here, we provide an evidence-based review of the pros and cons of commercially available automated methods for medical imaging diagnosis, including those based on artificial intelligence, to diagnose acute brain pathology on computed tomography or magnetic resonance imaging in patients with stroke.