Artificial Intelligence-Based Tool for Tumor Detection and Quantitative Tissue Analysis in Colorectal Specimens.

Digital pathology adoption allows for applying computational algorithms to routine pathology tasks. Our study aimed to develop a clinical-grade artificial intelligence (AI) tool for precise multiclass tissue segmentation in colorectal specimens (resections and biopsies) and clinically validate the tool for tumor detection in biopsy specimens. The training data set included 241 precisely manually annotated whole-slide images (WSIs) from multiple institutions. The algorithm was trained for semantic segmentation of 11 tissue classes with an additional module for biopsy WSI classification. Six case cohorts from 5 pathology departments (4 countries) were used for formal and clinical validation, digitized by 4 different scanning systems. The developed algorithm showed high precision of segmentation of different tissue classes in colorectal specimens with composite multiclass Dice score of up to 0.895 and pixel-wise tumor detection specificity and sensitivity of up to 0.958 and 0.987, respectively. In the clinical validation study on multiple external cohorts, the AI tool reached sensitivity of 1.0 and specificity of up to 0.969 for tumor detection in biopsy WSI. The AI tool analyzes most biopsy cases in less than 1 minute, allowing effective integration into clinical routine. We developed and extensively validated a highly accurate, clinical-grade tool for assistive diagnostic processing of colorectal specimens. This tool allows for quantitative deciphering of colorectal cancer tissue for development of prognostic and predictive biomarkers and personalization of oncologic care. This study is a foundation for a SemiCOL computational challenge. We open-source multiple manually annotated and weakly labeled test data sets, representing a significant contribution to the colorectal cancer computational pathology field.

Fuzzy association analysis for identifying climatic and socio-demographic factors impacting the spread of COVID-19.

Recently, the whole world witnessed the fatal outbreak of COVID-19 epidemic originating at Wuhan, Hubei province, China, during a mass gathering in a film festival. World Health Organization (WHO) has declared this COVID-19 as a pandemic due to its rapid spread across different countries within a few days. Several research works are being performed to understand the various influential factors responsible for spreading COVID. However, limited studies have been performed on how climatic and socio-demographic conditions may impact the spread of the virus. In this work, we aim to find the relationship of socio-demographic conditions, such as temperature, humidity, and population density of the regions, with the spread of COVID-19. The COVID data for different countries along with the social data are collected. For the experimental purpose, Fuzzy association rule mining is employed to infer the various relationships from the data. Moreover, to examine the seasonal effect, a streaming setting is also considered. The experimental results demonstrate various interesting insights to understand the impact of different factors on spreading COVID-19.

A network biology approach to identify crucial host targets for COVID-19.

The ongoing global pandemic of COVID-19, caused by SARS-CoV-2 has killed more than 5.9 million individuals out of ∼43 million confirmed infections. At present, several parts of the world are encountering the 3rd wave. Mass vaccination has been started in several countries but they are less likely to be broadly available for the current pandemic, repurposing of the existing drugs has drawn highest attention for an immediate solution. A recent publication has mapped the physical interactions of SARS-CoV-2 and human proteins by affinity-purification mass spectrometry (AP-MS) and identified 332 high-confidence SARS-CoV-2-human protein-protein interactions (PPIs). Here, we taken a network biology approach and constructed a human protein-protein interaction network (PPIN) with the above SARS-CoV-2 targeted proteins. We utilized a combination of essential network centrality measures and functional properties of the human proteins to identify the critical human targets of SARS-CoV-2. Four human proteins, namely PRKACA, RHOA, CDK5RAP2, and CEP250 have emerged as the best therapeutic targets, of which PRKACA and CEP250 were also found by another group as potential candidates for drug targets in COVID-19. We further found candidate drugs/compounds, such as guanosine triphosphate, remdesivir, adenosine monophosphate, MgATP, and H-89 dihydrochloride that bind the target human proteins. The urgency to prevent the spread of infection and the death of diseased individuals has prompted the search for agents from the pool of approved drugs to repurpose them for COVID-19. Our results indicate that host targeting therapy with the repurposed drugs may be a useful strategy for the treatment of SARS-CoV-2 infection.

Significance of an altered lncRNA landscape in schizophrenia and cognition: clues from a case-control association study.

Genetic etiology of schizophrenia is poorly understood despite large genome-wide association data. Long non-coding RNAs (lncRNAs) with a probable regulatory role are emerging as important players in neuro-psychiatric disorders including schizophrenia. Prioritising important lncRNAs and analyses of their holistic interaction with their target genes may provide insights into disease biology/etiology. Of the 3843 lncRNA SNPs reported in schizophrenia GWASs extracted using lincSNP 2.0, we prioritised n = 247 based on association strength, minor allele frequency and regulatory potential and mapped them to lncRNAs. lncRNAs were then prioritised based on their expression in brain using lncRBase, epigenetic role using 3D SNP and functional relevance to schizophrenia etiology. 18 SNPs were finally tested for association with schizophrenia (n = 930) and its endophenotypes-tardive dyskinesia (n = 176) and cognition (n = 565) using a case-control approach. Associated SNPs were characterised by ChIP seq, eQTL, and transcription factor binding site (TFBS) data using FeatSNP. Of the eight SNPs significantly associated, rs2072806 in lncRNA hsaLB_IO39983 with regulatory effect on BTN3A2 was associated with schizophrenia (p = 0.006); rs2710323 in hsaLB_IO_2331 with role in dysregulation of ITIH1 with tardive dyskinesia (p < 0.05); and four SNPs with significant cognition score reduction (p < 0.05) in cases. Two of these with two additional variants in eQTL were observed among controls (p < 0.05), acting likely as enhancer SNPs and/or altering TFBS of eQTL mapped downstream genes. This study highlights important lncRNAs in schizophrenia and provides a proof of concept of novel interactions of lncRNAs with protein-coding genes to elicit alterations in immune/inflammatory pathways of schizophrenia.

Generative Adversarial Networks: A Primer for Radiologists.

Artificial intelligence techniques involving the use of artificial neural networks-that is, deep learning techniques-are expected to have a major effect on radiology. Some of the most exciting applications of deep learning in radiology make use of generative adversarial networks (GANs). GANs consist of two artificial neural networks that are jointly optimized but with opposing goals. One neural network, the generator, aims to synthesize images that cannot be distinguished from real images. The second neural network, the discriminator, aims to distinguish these synthetic images from real images. These deep learning models allow, among other applications, the synthesis of new images, acceleration of image acquisitions, reduction of imaging artifacts, efficient and accurate conversion between medical images acquired with different modalities, and identification of abnormalities depicted on images. The authors provide an introduction to GANs and adversarial deep learning methods. In addition, the different ways in which GANs can be used for image synthesis and image-to-image translation tasks, as well as the principles underlying conditional GANs and cycle-consistent GANs, are described. Illustrated examples of GAN applications in radiologic image analysis for different imaging modalities and different tasks are provided. The clinical potential of GANs, future clinical GAN applications, and potential pitfalls and caveats that radiologists should be aware of also are discussed in this review. The online slide presentation from the RSNA Annual Meeting is available for this article. ©RSNA, 2021.

The unhealthy = tasty belief is associated with BMI through reduced consumption of vegetables: A cross-national and mediational analysis.

Obesity is one of the greatest public health challenges of modern times and its prevalence is increasing worldwide. With food so abundant in developed countries, many people face a conflict between desires for short-term taste and the goal of long-term health, multiple times a day. Recent research suggests that consumers often resolve these conflicts based on their lay beliefs about the healthiness and tastiness of food. Consequently, such lay beliefs can play critical roles not just in food choice but also weight gain. In this research, we show, across six countries and through mediation analysis, that adults who believe that tasty food is unhealthy (the Unhealthy = Tasty Intuition, or "UTI"; Raghunathan, Naylor, & Hoyer 2006) are less likely to consume healthy food, and thereby have a higher body mass index (BMI). In Study 1, we conducted a cross-sectional survey in five countries (Australia, Germany, Hong Kong, India, and the UK), and found that greater strength of belief in UTI was associated with higher BMI, and this relationship was mediated by lower consumption of fruits and vegetables. The observed patterns largely converged across the sampled Western and Asian-Pacific countries. In Study 2, we teased apart the mediating role of vegetable versus fruit consumption and also addressed the issue of reversed causality by predicting BMI with a measure of UTI belief taken 30 months previously. We found that vegetable consumption, but not fruit consumption, mediated the association between UTI belief and BMI. Our findings contribute to the literature by showing how lay beliefs about food can have pervasive and long-lasting effects on dietary practices and health worldwide. Implications for public policy and health practitioners are discussed.

Identification of infectious disease-associated host genes using machine learning techniques.

BACKGROUND: With the global spread of multidrug resistance in pathogenic microbes, infectious diseases emerge as a key public health concern of the recent time. Identification of host genes associated with infectious diseases will improve our understanding about the mechanisms behind their development and help to identify novel therapeutic targets. RESULTS: We developed a machine learning techniques-based classification approach to identify infectious disease-associated host genes by integrating sequence and protein interaction network features. Among different methods, Deep Neural Networks (DNN) model with 16 selected features for pseudo-amino acid composition (PAAC) and network properties achieved the highest accuracy of 86.33% with sensitivity of 85.61% and specificity of 86.57%. The DNN classifier also attained an accuracy of 83.33% on a blind dataset and a sensitivity of 83.1% on an independent dataset. Furthermore, to predict unknown infectious disease-associated host genes, we applied the proposed DNN model to all reviewed proteins from the database. Seventy-six out of 100 highly-predicted infectious disease-associated genes from our study were also found in experimentally-verified human-pathogen protein-protein interactions (PPIs). Finally, we validated the highly-predicted infectious disease-associated genes by disease and gene ontology enrichment analysis and found that many of them are shared by one or more of the other diseases, such as cancer, metabolic and immune related diseases. CONCLUSIONS: To the best of our knowledge, this is the first computational method to identify infectious disease-associated host genes. The proposed method will help large-scale prediction of host genes associated with infectious-diseases. However, our results indicated that for small datasets, advanced DNN-based method does not offer significant advantage over the simpler supervised machine learning techniques, such as Support Vector Machine (SVM) or Random Forest (RF) for the prediction of infectious disease-associated host genes. Significant overlap of infectious disease with cancer and metabolic disease on disease and gene ontology enrichment analysis suggests that these diseases perturb the functions of the same cellular signaling pathways and may be treated by drugs that tend to reverse these perturbations. Moreover, identification of novel candidate genes associated with infectious diseases would help us to explain disease pathogenesis further and develop novel therapeutics.

A comprehensive analysis on preservation patterns of gene co-expression networks during Alzheimer's disease progression.

BACKGROUND: Alzheimer's disease (AD) is a chronic neuro-degenerative disruption of the brain which involves in large scale transcriptomic variation. The disease does not impact every regions of the brain at the same time, instead it progresses slowly involving somewhat sequential interaction with different regions. Analysis of the expression patterns of the genes in different regions of the brain influenced in AD surely contribute for a enhanced comprehension of AD pathogenesis and shed light on the early characterization of the disease. RESULTS: Here, we have proposed a framework to identify perturbation and preservation characteristics of gene expression patterns across six distinct regions of the brain ("EC", "HIP", "PC", "MTG", "SFG", and "VCX") affected in AD. Co-expression modules were discovered considering a couple of regions at once. These are then analyzed to know the preservation and perturbation characteristics. Different module preservation statistics and a rank aggregation mechanism have been adopted to detect the changes of expression patterns across brain regions. Gene ontology (GO) and pathway based analysis were also carried out to know the biological meaning of preserved and perturbed modules. CONCLUSIONS: In this article, we have extensively studied the preservation patterns of co-expressed modules in six distinct brain regions affected in AD. Some modules are emerged as the most preserved while some others are detected as perturbed between a pair of brain regions. Further investigation on the topological properties of preserved and non-preserved modules reveals a substantial association amongst "betweenness centrality" and "degree" of the involved genes. Our findings may render a deeper realization of the preservation characteristics of gene expression patterns in discrete brain regions affected by AD.

Preservation affinity in consensus modules among stages of HIV-1 progression.

BACKGROUND: Analysis of gene expression data provides valuable insights into disease mechanism. Investigating relationship among co-expression modules of different stages is a meaningful tool to understand the way in which a disease progresses. Identifying topological preservation of modular structure also contributes to that understanding. METHODS: HIV-1 disease provides a well-documented progression pattern through three stages of infection: acute, chronic and non-progressor. In this article, we have developed a novel framework to describe the relationship among the consensus (or shared) co-expression modules for each pair of HIV-1 infection stages. The consensus modules are identified to assess the preservation of network properties. We have investigated the preservation patterns of co-expression networks during HIV-1 disease progression through an eigengene-based approach. RESULTS: We discovered that the expression patterns of consensus modules have a strong preservation during the transitions of three infection stages. In particular, it is noticed that between acute and non-progressor stages the preservation is slightly more than the other pair of stages. Moreover, we have constructed eigengene networks for the identified consensus modules and observed the preservation structure among them. Some consensus modules are marked as preserved in two pairs of stages and are analyzed further to form a higher order meta-network consisting of a group of preserved modules. Additionally, we observed that module membership (MM) values of genes within a module are consistent with the preservation characteristics. The MM values of genes within a pair of preserved modules show strong correlation patterns across two infection stages. CONCLUSIONS: We have performed an extensive analysis to discover preservation pattern of co-expression network constructed from microarray gene expression data of three different HIV-1 progression stages. The preservation pattern is investigated through identification of consensus modules in each pair of infection stages. It is observed that the preservation of the expression pattern of consensus modules remains more prominent during the transition of infection from acute stage to non-progressor stage. Additionally, we observed that the module membership values of genes are coherent with preserved modules across the HIV-1 progression stages.

Characterization of SNPs in the dopamine-ß-hydroxylase gene providing new insights into its structure-function relationship.

Dopamine-ß-hydroxylase (DBH, EC 1.14.17.1), an oxido-reductase that catalyses the conversion of dopamine to norepinephrine, is largely expressed in sympathetic neurons and adrenal medulla. Several regulatory and structural variants in DBH associated with various neuropsychiatric, cardiovascular diseases and a few that may determine enzyme activity have also been identified. Due to paucity of studies on functional characterization of DBH variants, its structure-function relationship is poorly understood. The purpose of the study was to characterize five non-synonymous (ns) variants that were prioritized either based on previous association studies or Sorting Tolerant From Intolerant (SIFT) algorithm. The DBH ORF with wild type (WT) and site-directed mutagenized variants were transfected into HEK293 cells to generate transient and stable lines expressing these variant enzymes. Activity was determined by UPLC-PDA and corresponding quantity by MRMHR on a TripleTOF 5600 MS respectively of spent media from stable cell lines. Homospecific activity computed for the WT and variant proteins showed a marginal decrease in A318S, W544S and R549C variants. In transient cell lines, differential secretion was observed in the case of L317P, W544S and R549C. Secretory defect in L317P was confirmed by localization in ER. R549C exhibited both decreased homospecific activity and differential secretion. Of note, all the variants were seen to be destabilizing based on in silico folding analysis and molecular dynamics (MD) simulation, lending support to our experimental observations. These novel genotype-phenotype correlations in this gene of considerable pharmacological relevance have implications for dopamine-related disorders.

A review of in silico approaches for analysis and prediction of HIV-1-human protein-protein interactions.

The computational or in silico approaches for analysing the HIV-1-human protein-protein interaction (PPI) network, predicting different host cellular factors and PPIs and discovering several pathways are gaining popularity in the field of HIV research. Although there exist quite a few studies in this regard, no previous effort has been made to review these works in a comprehensive manner. Here we review the computational approaches that are devoted to the analysis and prediction of HIV-1-human PPIs. We have broadly categorized these studies into two fields: computational analysis of HIV-1-human PPI network and prediction of novel PPIs. We have also presented a comparative assessment of these studies and proposed some methodologies for discussing the implication of their results. We have also reviewed different computational techniques for predicting HIV-1-human PPIs and provided a comparative study of their applicability. We believe that our effort will provide helpful insights to the HIV research community.

Multiobjective triclustering of time-series transcriptome data reveals key genes of biological processes.

BACKGROUND: Exploratory analysis of multi-dimensional high-throughput datasets, such as microarray gene expression time series, may be instrumental in understanding the genetic programs underlying numerous biological processes. In such datasets, variations in the gene expression profiles are usually observed across replicates and time points. Thus mining the temporal expression patterns in such multi-dimensional datasets may not only provide insights into the key biological processes governing organs to grow and develop but also facilitate the understanding of the underlying complex gene regulatory circuits. RESULTS: In this work we have developed an evolutionary multi-objective optimization for our previously introduced triclustering algorithm δ-TRIMAX. Its aim is to make optimal use of δ-TRIMAX in extracting groups of co-expressed genes from time series gene expression data, or from any 3D gene expression dataset, by adding the powerful capabilities of an evolutionary algorithm to retrieve overlapping triclusters. We have compared the performance of our newly developed algorithm, EMOA- δ-TRIMAX, with that of other existing triclustering approaches using four artificial dataset and three real-life datasets. Moreover, we have analyzed the results of our algorithm on one of these real-life datasets monitoring the differentiation of human induced pluripotent stem cells (hiPSC) into mature cardiomyocytes. For each group of co-expressed genes belonging to one tricluster, we identified key genes by computing their membership values within the tricluster. It turned out that to a very high percentage, these key genes were significantly enriched in Gene Ontology categories or KEGG pathways that fitted very well to the biological context of cardiomyocytes differentiation. CONCLUSIONS: EMOA- δ-TRIMAX has proven instrumental in identifying groups of genes in transcriptomic data sets that represent the functional categories constituting the biological process under study. The executable file can be found at http://www.bioinf.med.uni-goettingen.de/fileadmin/download/EMOA-delta-TRIMAX.tar.gz .

Implications of agricultural land use change to ecosystem services in the Ganges delta.

Ecosystems provide the basis for human civilization and natural capital for green economy and sustainable development. Ecosystem services may range from crops, fish, freshwater to those that are harder to see such as erosion regulation, carbon sequestration, and pest control. Land use changes have been identified as the main sources of coastal and marine pollution in Bangladesh. This paper explores the temporal variation of agricultural land use change and its implications with ecosystem services in the Ganges delta. With time agricultural lands have been decreased and wetlands have been increased at a very high rate mainly due to the growing popularity of saltwater shrimp farming. In a span of 28 years, the agricultural lands have been reduced by approximately 50%, while the wetlands have been increased by over 500%. A large portion (nearly 40%) of the study area is covered by the Sundarbans which remained almost constant which can be attributed to the strict regulatory intervention to preserve the Sundarbans. The settlement & others land use type has also been increased to nearly 5%. There is a gradual uptrend of shrimp and fish production in the study area. The findings suggest that there are significant linkages between agricultural land use change and ecosystem services in the Ganges delta in Bangladesh. The continuous decline of agricultural land (due to salinization) and an increase of wetland have been attributed to the conversion of agricultural land into shrimp farming in the study area. Such land use change requires significant capital, therefore, only investors and wealthier land owners can get the higher profit from the land conversion while the poor people is left with the environmental consequences that affect their long-term lives and livelihood. An environmental management plan is proposed for sustainable land use in the Ganges delta in Bangladesh.

Incorporating the type and direction information in predicting novel regulatory interactions between HIV-1 and human proteins using a biclustering approach.

BACKGROUND: Discovering novel interactions between HIV-1 and human proteins would greatly contribute to different areas of HIV research. Identification of such interactions leads to a greater insight into drug target prediction. Some recent studies have been conducted for computational prediction of new interactions based on the experimentally validated information stored in a HIV-1-human protein-protein interaction database. However, these techniques do not predict any regulatory mechanism between HIV-1 and human proteins by considering interaction types and direction of regulation of interactions. RESULTS: Here we present an association rule mining technique based on biclustering for discovering a set of rules among human and HIV-1 proteins using the publicly available HIV-1-human PPI database. These rules are subsequently utilized to predict some novel interactions among HIV-1 and human proteins. For prediction purpose both the interaction types and direction of regulation of interactions, (i.e., virus-to-host or host-to-virus) are considered here to provide important additional information about the regulation pattern of interactions. We have also studied the biclusters and analyzed the significant GO terms and KEGG pathways in which the human proteins of the biclusters participate. Moreover the predicted rules have also been analyzed to discover regulatory relationship between some human proteins in course of HIV-1 infection. Some experimental evidences of our predicted interactions have been found by searching the recent literatures in PUBMED. We have also highlighted some human proteins that are likely to act against the HIV-1 attack. CONCLUSIONS: We pose the problem of identifying new regulatory interactions between HIV-1 and human proteins based on the existing PPI database as an association rule mining problem based on biclustering algorithm. We discover some novel regulatory interactions between HIV-1 and human proteins. Significant number of predicted interactions has been found to be supported by recent literature.

A comprehensive review of machine learning techniques for multi-omics data integration: challenges and applications in precision oncology.

Multi-omics data play a crucial role in precision medicine, mainly to understand the diverse biological interaction between different omics. Machine learning approaches have been extensively employed in this context over the years. This review aims to comprehensively summarize and categorize these advancements, focusing on the integration of multi-omics data, which includes genomics, transcriptomics, proteomics and metabolomics, alongside clinical data. We discuss various machine learning techniques and computational methodologies used for integrating distinct omics datasets and provide valuable insights into their application. The review emphasizes both the challenges and opportunities present in multi-omics data integration, precision medicine and patient stratification, offering practical recommendations for method selection in various scenarios. Recent advances in deep learning and network-based approaches are also explored, highlighting their potential to harmonize diverse biological information layers. Additionally, we present a roadmap for the integration of multi-omics data in precision oncology, outlining the advantages, challenges and implementation difficulties. Hence this review offers a thorough overview of current literature, providing researchers with insights into machine learning techniques for patient stratification, particularly in precision oncology. Contact:  anirban@klyuniv.ac.in.

A randomized optimal k-mer indexing approach for efficient parallel genome sequence compression.

Next Generation Sequencing (NGS) technology generates massive amounts of genome sequence that increases rapidly over time. As a result, there is a growing need for efficient compression algorithms to facilitate the processing, storage, transmission, and analysis of large-scale genome sequences. Over the past 31 years, numerous state-of-the-art compression algorithms have been developed. The performance of any compression algorithm is measured by three main compression metrics: compression ratio, time, and memory usage. Existing k-mer hash indexing systems take more time, due to the decision-making process based on compression results. In this paper, we propose a two-phase reference genome compression algorithm using optimal k-mer length (RGCOK). Reference-based compression takes advantage of the inter-similarity between chromosomes of the same species. RGCOK achieves this by finding the optimal k-mer length for matching, using a randomization method and hashing. The performance of RGCOK was evaluated on three different benchmark data sets: novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), Homo sapiens, and other species sequences using an Amazon AWS virtual cloud machine. Experiments showed that the optimal k-mer finding time by RGCOK is around 45.28 min, whereas the time for existing state-of-the-art algorithms HiRGC, SCCG, and HRCM ranges from 58 min to 8.97 h.

Dynamics of self-control during choice and post-choice consumption quantity.

Observed choices between options representing a relative vice and a relative virtue have commonly been used as a measure of eating self-control in the literature. However, even though self-control operations may manifest across the post-choice consumption stage, either similarly or in different ways from the choice stage, most prior research has ignored consumption quantity of the chosen option. While the behavior of choosing a virtue instead of a vice does manifest self-control, we examine how this plays out in post-choice consumption. Specifically, we find that when processing resources are limited, after having chosen a virtue food, unrestrained eaters ironically consumed greater quantities and therefore more calories than restrained eaters (Study 1). This reflects more persistent self-control in the post-choice consumption stage among restrained eaters than unrestrained eaters, and occurs because choosing a virtue lowers accessibility of the self-control goal among unrestrained eaters relative to restrained eaters (Study 2), thereby increasing intake of the virtuous food. In contrast, subsequent to having chosen a vice, unrestrained eaters and restrained eaters did not show any such difference in intake (Study 1) or goal accessibility (Study 2). Together, these results reveal that persistence of self-control in the post-choice consumption stage depends on individuals' dietary restraint and their initial exercise of self-control in the choice decision. The mere act of choosing a virtue satisfies unrestrained eaters' self-control goal and leads to increased food intake, whereas the same act keeps the same goal activated among restrained eaters who reduce intake of the chosen virtue. Put differently, persistent self-control across choice and quantity decisions is observed only when those with a dietary goal show successful self-control enactment in the choice stage. We therefore highlight that the operation of self-control can be dynamic within a consumption episode, and thus, choice and post-choice quantity are both informative of self-control.

Sliding Window Optimal Transport for Open World Artifact Detection in Histopathology.

Histological images are frequently impaired by local artifacts from scanner malfunctions or iatrogenic processes - caused by preparation - impacting the performance of Deep Learning models. Models often struggle with the slightest out-of-distribution shifts, resulting in compromised performance. Detecting artifacts and failure modes of the models is crucial to ensure open-world applicability to whole slide images for tasks like segmentation or diagnosis. We introduce a novel technique for out-of-distribution detection within whole slide images, compatible with any segmentation or classification model. Our approach tiles multi-layer features into sliding window patches and leverages optimal transport to align them with recognized in-distribution samples. We average the optimal transport costs over tiles and layers to detect out-of-distribution samples. Notably, our method excels in identifying failure modes that would harm downstream performance, surpassing contemporary out-of-distribution detection techniques. We evaluate our method for both natural and synthetic artifacts, considering distribution shifts of various sizes and types. The results confirm that our technique outperforms alternative methods for artifact detection. We assess our method components and the ability to negate the impact of artifacts on the downstream tasks. Finally, we demonstrate that our method can mitigate the risk of performance drops in downstream tasks, enhancing reliability by up to 77%. In testing 7 annotated whole slide images with natural artifacts, our method boosted the Dice score by 68%, highlighting its real open-world utility.

Closing the loop for AI-ready radiology.

BACKGROUND: In recent years, AI has made significant advancements in medical diagnosis and prognosis. However, the incorporation of AI into clinical practice is still challenging and under-appreciated. We aim to demonstrate a possible vertical integration approach to close the loop for AI-ready radiology. METHOD: This study highlights the importance of two-way communication for AI-assisted radiology. As a key part of the methodology, it demonstrates the integration of AI systems into clinical practice with structured reports and AI visualization, giving more insight into the AI system. By integrating cooperative lifelong learning into the AI system, we ensure the long-term effectiveness of the AI system, while keeping the radiologist in the loop.  RESULTS: We demonstrate the use of lifelong learning for AI systems by incorporating AI visualization and structured reports. We evaluate Memory Aware-Synapses and Rehearsal approach and find that both approaches work in practice. Furthermore, we see the advantage of lifelong learning algorithms that do not require the storing or maintaining of samples from previous datasets. CONCLUSION: In conclusion, incorporating AI into the clinical routine of radiology requires a two-way communication approach and seamless integration of the AI system, which we achieve with structured reports and visualization of the insight gained by the model. Closing the loop for radiology leads to successful integration, enabling lifelong learning for the AI system, which is crucial for sustainable long-term performance. KEY POINTS: · The integration of AI systems into the clinical routine with structured reports and AI visualization.. · Two-way communication between AI and radiologists is necessary to enable AI that keeps the radiologist in the loop.. · Closing the loop enables lifelong learning, which is crucial for long-term, high-performing AI in radiology..

Comparison of Ultra-High-Resolution and Normal-Resolution CT-Angiography for Intracranial Aneurysm Detection in Patients with Subarachnoid Hemorrhage.

RATIONALE AND OBJECTIVES: Ruptured intracranial aneurysms (IAs) are the leading cause for atraumatic subarachnoid hemorrhage. In case of aneurysm rupture, patients may face life-threatening complications and require aneurysm occlusion. Detection of the aneurysm in computed tomography (CT) imaging is therefore essential for patient outcome. This study provides an evaluation of the diagnostic accuracy of Ultra-High-Resolution Computed Tomography Angiography (UHR-CTA) and Normal-Resolution Computed Tomography Angiography (NR-CTA) concerning IA detection and characterization. MATERIALS AND METHODS: Consecutive patients with atraumatic subarachnoid hemorrhage who received Digital Subtraction Angiography (DSA) and either UHR-CTA or NR-CTA were retrospectively included. Three readers evaluated CT-Angiography regarding image quality, diagnostic confidence and presence of IAs. Sensitivity and specificity were calculated on patient-level and segment-level with reference standard DSA-imaging. CTA patient radiation exposure (effective dose) was compared. RESULTS: One hundred and eight patients were identified (mean age = 57.8 ±â€¯14.1 years, 65 women). UHR-CTA revealed significantly higher image quality and diagnostic confidence (P < 0.001) for all readers and significantly lower effective dose (P < 0.001). Readers correctly classified ≥55/56 patients on UHR-CTA and ≥44/52 patients on NR-CTA. We noted significantly higher patient-level sensitivity for UHR-CTA compared to NR-CTA for all three readers (reader 1: 41/41 [100%] vs. 28/34 [82%], reader 2: 41/41 [100%] vs. 30/34 [88%], reader 3: 41/41 [100%] vs. 30/34 [88%], P ≤ 0.04). Segment-level analysis also revealed significantly higher sensitivity for UHR-CTA compared to NR-CTA for all three readers (reader 1: 47/49 [96%] vs. 34/45 [76%], reader 2: 47/49 [96%] vs. 37/45 [82%], reader 3: 48/49 [98%] vs. 37/45 [82%], P ≤ 0.04). Specificity was comparable for both techniques. CONCLUSION: We found Ultra-High-Resolution CT-Angiography to provide higher sensitivity than Normal-Resolution CT-Angiography for the detection of intracranial aneurysms in patients with aneurysmal subarachnoid hemorrhage while improving image quality and reducing patient radiation exposure.