Using histopathology latent diffusion models as privacy-preserving dataset augmenters improves downstream classification performance.

Latent diffusion models (LDMs) have emerged as a state-of-the-art image generation method, outperforming previous Generative Adversarial Networks (GANs) in terms of training stability and image quality. In computational pathology, generative models are valuable for data sharing and data augmentation. However, the impact of LDM-generated images on histopathology tasks compared to traditional GANs has not been systematically studied. We trained three LDMs and a styleGAN2 model on histology tiles from nine colorectal cancer (CRC) tissue classes. The LDMs include 1) a fine-tuned version of stable diffusion v1.4, 2) a Kullback-Leibler (KL)-autoencoder (KLF8-DM), and 3) a vector quantized (VQ)-autoencoder deploying LDM (VQF8-DM). We assessed image quality through expert ratings, dimensional reduction methods, distribution similarity measures, and their impact on training a multiclass tissue classifier. Additionally, we investigated image memorization in the KLF8-DM and styleGAN2 models. All models provided a high image quality, with the KLF8-DM achieving the best Frechet Inception Distance (FID) and expert rating scores for complex tissue classes. For simpler classes, the VQF8-DM and styleGAN2 models performed better. Image memorization was negligible for both styleGAN2 and KLF8-DM models. Classifiers trained on a mix of KLF8-DM generated and real images achieved a 4% improvement in overall classification accuracy, highlighting the usefulness of these images for dataset augmentation. Our systematic study of generative methods showed that KLF8-DM produces the highest quality images with negligible image memorization. The higher classifier performance in the generatively augmented dataset suggests that this augmentation technique can be employed to enhance histopathology classifiers for various tasks.

Built to Last? Reproducibility and Reusability of Deep Learning Algorithms in Computational Pathology.

Recent progress in computational pathology has been driven by deep learning. While code and data availability are essential to reproduce findings from preceding publications, ensuring a deep learning model's reusability is more challenging. For that, the codebase should be well-documented and easy to integrate into existing workflows and models should be robust toward noise and generalizable toward data from different sources. Strikingly, only a few computational pathology algorithms have been reused by other researchers so far, let alone employed in a clinical setting. To assess the current state of reproducibility and reusability of computational pathology algorithms, we evaluated peer-reviewed articles available in PubMed, published between January 2019 and March 2021, in 5 use cases: stain normalization; tissue type segmentation; evaluation of cell-level features; genetic alteration prediction; and inference of grading, staging, and prognostic information. We compiled criteria for data and code availability and statistical result analysis and assessed them in 160 publications. We found that only one-quarter (41 of 160 publications) made code publicly available. Among these 41 studies, three-quarters (30 of 41) analyzed their results statistically, half of them (20 of 41) released their trained model weights, and approximately a third (16 of 41) used an independent cohort for evaluation. Our review is intended for both pathologists interested in deep learning and researchers applying algorithms to computational pathology challenges. We provide a detailed overview of publications with published code in the field, list reusable data handling tools, and provide criteria for reproducibility and reusability.

The future landscape of large language models in medicine.

Large language models (LLMs) are artificial intelligence (AI) tools specifically trained to process and generate text. LLMs attracted substantial public attention after OpenAI's ChatGPT was made publicly available in November 2022. LLMs can often answer questions, summarize, paraphrase and translate text on a level that is nearly indistinguishable from human capabilities. The possibility to actively interact with models like ChatGPT makes LLMs attractive tools in various fields, including medicine. While these models have the potential to democratize medical knowledge and facilitate access to healthcare, they could equally distribute misinformation and exacerbate scientific misconduct due to a lack of accountability and transparency. In this article, we provide a systematic and comprehensive overview of the potentials and limitations of LLMs in clinical practice, medical research and medical education.

Transformer-based biomarker prediction from colorectal cancer histology: A large-scale multicentric study.

Deep learning (DL) can accelerate the prediction of prognostic biomarkers from routine pathology slides in colorectal cancer (CRC). However, current approaches rely on convolutional neural networks (CNNs) and have mostly been validated on small patient cohorts. Here, we develop a new transformer-based pipeline for end-to-end biomarker prediction from pathology slides by combining a pre-trained transformer encoder with a transformer network for patch aggregation. Our transformer-based approach substantially improves the performance, generalizability, data efficiency, and interpretability as compared with current state-of-the-art algorithms. After training and evaluating on a large multicenter cohort of over 13,000 patients from 16 colorectal cancer cohorts, we achieve a sensitivity of 0.99 with a negative predictive value of over 0.99 for prediction of microsatellite instability (MSI) on surgical resection specimens. We demonstrate that resection specimen-only training reaches clinical-grade performance on endoscopic biopsy tissue, solving a long-standing diagnostic problem.

Multi-Modality Microscopy Image Style Augmentation for Nuclei Segmentation.

Annotating microscopy images for nuclei segmentation by medical experts is laborious and time-consuming. To leverage the few existing annotations, also across multiple modalities, we propose a novel microscopy-style augmentation technique based on a generative adversarial network (GAN). Unlike other style transfer methods, it can not only deal with different cell assay types and lighting conditions, but also with different imaging modalities, such as bright-field and fluorescence microscopy. Using disentangled representations for content and style, we can preserve the structure of the original image while altering its style during augmentation. We evaluate our data augmentation on the 2018 Data Science Bowl dataset consisting of various cell assays, lighting conditions, and imaging modalities. With our style augmentation, the segmentation accuracy of the two top-ranked Mask R-CNN-based nuclei segmentation algorithms in the competition increases significantly. Thus, our augmentation technique renders the downstream task more robust to the test data heterogeneity and helps counteract class imbalance without resampling of minority classes.

Feasibility of transcutaneous auricular vagus nerve stimulation in treatment of drug resistant epilepsy: A multicenter prospective study.

BACKGROUND: Transcutaneous auricular vagus nerve stimulation (ta-VNS) is a new non-invasive technique developed as treatment option for drug resistant epilepsy. A few studies have been carried out showing that the efficacy and tolerability of ta-VNS is comparable with traditional implanted VNS but the feasibility of the therapy has been poorly described. This study aimed to explore potential clinical benefits of ta-VNS and to evaluate adaptation, compliance, as well as the usability of the device from a service design perspective. METHODS: A prospective, multicenter, clinical, investigator-initiated trial was conducted using the NEMOS® ta-VNS device. After eight weeks baseline, all subjects started ta-VNS with individually adjusted currents for four hours per day for six-months (first endpoint) followed by optional 12 months follow-up (second endpoint). The primary outcome was six months retention rate of ta-VNS therapy. Secondary outcomes included the user retention rate at 12 months follow-up, compliance, changes in scores of psychometric measures. For the study of feasibility, a service design questionnaire on medical devices used in the home was developed. RESULTS: In total 37 subjects had been included in the study after 45 months where the study was prematurely terminated due to recruitment problems and due to a high drop-out rate. Twenty-two subjects (59 %) completed the first six months of the study and in total six subjects (16 %) completed the following 12 months follow-up. The reasons for discontinuation were a mixture of medical and practical issues of which the majority were related to a combination of both. Those, who managed to continue to use ta-VNS throughout the study, gave generally higher scores for the device usability and compatibility with lifestyle. The study turned out to be inadequately powered to reach any conclusion in terms of the clinical benefits of ta-VNS but present an example of difficulties that are encountered in conducting high-quality studies with digital devices. CONCLUSION: The feasibility of ta-VNS therapy showed to be relatively modest which is most likely due to practical usability issues and lifestyle fits. The results of this study stress the importance of generating data based on patients experiences at an early stage during the development phase and when designing clinical trials on medical devices that depend on patient's active participation and motivation.

Antagonistic species interaction drives selection for sex in a predator-prey system.

The evolutionary maintenance of sexual reproduction has long challenged biologists as the majority of species reproduce sexually despite inherent costs. Providing a general explanation for the evolutionary success of sex has thus proven difficult and resulted in numerous hypotheses. A leading hypothesis suggests that antagonistic species interaction can generate conditions selecting for increased sex due to the production of rare or novel genotypes that are beneficial for rapid adaptation to recurrent environmental change brought on by antagonism. To test this ecology-based hypothesis, we conducted experimental evolution in a predator (rotifer)-prey (algal) system by using continuous cultures to track predator-prey dynamics and in situ rates of sex in the prey over time and within replicated experimental populations. Overall, we found that predator-mediated fluctuating selection for competitive versus defended prey resulted in higher rates of genetic mixing in the prey. More specifically, our results showed that fluctuating population sizes of predator and prey, coupled with a trade-off in the prey, drove the sort of recurrent environmental change that could provide a benefit to sex in the prey, despite inherent costs. We end with a discussion of potential population genetic mechanisms underlying increased selection for sex in this system, based on our application of a general theoretical framework for measuring the effects of sex over time, and interpreting how these effects can lead to inferences about the conditions selecting for or against sexual reproduction in a system with antagonistic species interaction.

Predictors of haemodynamic structural valve deterioration following transcatheter aortic valve implantation with latest-generation balloon-expandable valves.

AIMS: Elevated gradients have been proposed to be associated with haemodynamic structural valve deterioration (SVD) after transcatheter aortic valve implantation (TAVI); however, data regarding their characterisation remain scarce. This study sought to investigate the prevalence and predictors of moderate or greater SVD and the prevalence of valve thrombosis during follow-up after TAVI with balloon-expandable valves. METHODS AND RESULTS: A total of 691 patients undergoing transfemoral TAVI were enrolled. The primary endpoint was moderate or severe haemodynamic SVD during 12-month follow-up after TAVI, defined as (I) mean transvalvular gradient ≥20 mmHg or (II) mean transvalvular gradient increase ≥10 mmHg. The primary endpoint was observed in 10.3% after TAVI. Use of a 20 mm valve, valve-in-valve procedure and oral anticoagulation (OAC) were independently associated with haemodynamic SVD, whereas valve-in-valve procedure and OAC were the only significant variables after accounting for death as a competing event. OAC was significantly associated with haemodynamic SVD (RR 8.65; p=0.004) and death (RR 3.57; p=0.06), whereas valve-in-valve procedure was only associated with haemodynamic SVD (RR 52.76; p<0.001). Valve thrombosis was present in 0.87% (6/691) of the patients. CONCLUSIONS: The prevalence of moderate or greater haemodynamic SVD during the first 12 months after TAVI is 10.3%. Procedural factors and pharmacotherapy seem to play a key role during manifestation. Future studies should focus on the underlying mechanisms.

Cryptic changes in immune response and fitness in Tribolium castaneum as a consequence of coevolution with Beauveria bassiana.

Immunity is a key trait in host defence against parasites and is thus likely to be under selection during host-parasite coevolution. Broadly, the immune system consists of several lines of defence including physiological innate immunity, physical barriers such as the cuticle, avoidance behaviours and in some cases antimicrobial secretions. The defence conferring the highest fitness benefit may be situation specific and depend on the taxon and infection route of the parasite. We carried out a host-parasite coevolution experiment between the red flour beetle T. castaneum, which possesses a comprehensive immune system including the ability to secrete antimicrobial compounds into its environment, and the generalist entomopathogenic fungus Beauveria bassiana. We measured levels of external immunity (benzoquinone secretion) and an internal immune trait, phenoloxidase (PO) activity throughout and in F2 to beetles at the end of the experiment. Survival (a proxy for resistance) of F2 coevolved and control beetles exposed to the fungus was also measured. No change in external immunity or survival was observed as a consequence of host-parasite coevolution, however, PO responses in evolved beetles showed increased flexibility dependent on the route of infection of the parasite. This more flexible PO response appeared to result in beetle populations being better able to cope with the parasite, buffering their fitness during the course of the coevolution experiment. This represents a subtle but significant adaptation to the presence of a parasite over evolutionary time.

The native microbiome of the nematode Caenorhabditis elegans: gateway to a new host-microbiome model.

BACKGROUND: Host-microbe associations underlie many key processes of host development, immunity, and life history. Yet, none of the current research on the central model species Caenorhabditis elegans considers the worm's natural microbiome. Instead, almost all laboratories exclusively use the canonical strain N2 and derived mutants, maintained through routine bleach sterilization in monoxenic cultures with an E. coli strain as food. Here, we characterize for the first time the native microbiome of C. elegans and assess its influence on nematode life history characteristics. RESULTS: Nematodes sampled directly from their native habitats carry a species-rich bacterial community, dominated by Proteobacteria such as Enterobacteriaceae and members of the genera Pseudomonas, Stenotrophomonas, Ochrobactrum, and Sphingomonas. The C. elegans microbiome is distinct from that of the worm's natural environment and the congeneric species C. remanei. Exposure to a derived experimental microbiome revealed that bacterial composition is influenced by host developmental stage and genotype. These experiments also showed that the microbes enhance host fitness under standard and also stressful conditions (e.g., high temperature and either low or high osmolarity). Taking advantage of the nematode's transparency, we further demonstrate that several Proteobacteria are able to enter the C. elegans gut and that an Ochrobactrum isolate even seems to be able to persist in the intestines under stressful conditions. Moreover, three Pseudomonas isolates produce an anti-fungal effect in vitro which we show can contribute to the worm's defense against fungal pathogens in vivo. CONCLUSION: This first systematic analysis of the nematode's native microbiome reveals a species-rich bacterial community to be associated with C. elegans, which is likely of central importance for our understanding of the worm's biology. The information acquired and the microbial isolates now available for experimental work establishes C. elegans as a tractable model for the in-depth dissection of host-microbiome interactions.