TimeREISE: Time Series Randomized Evolving Input Sample Explanation.

Deep neural networks are one of the most successful classifiers across different domains. However, their use is limited in safety-critical areas due to their limitations concerning interpretability. The research field of explainable artificial intelligence addresses this problem. However, most interpretability methods align to the imaging modality by design. The paper introduces TimeREISE, a model agnostic attribution method that shows success in the context of time series classification. The method applies perturbations to the input and considers different attribution map characteristics such as the granularity and density of an attribution map. The approach demonstrates superior performance compared to existing methods concerning different well-established measurements. TimeREISE shows impressive results in the deletion and insertion test, Infidelity, and Sensitivity. Concerning the continuity of an explanation, it showed superior performance while preserving the correctness of the attribution map. Additional sanity checks prove the correctness of the approach and its dependency on the model parameters. TimeREISE scales well with an increasing number of channels and timesteps. TimeREISE applies to any time series classification network and does not rely on prior data knowledge. TimeREISE is suited for any usecase independent of dataset characteristics such as sequence length, channel number, and number of classes.

TSInsight: A Local-Global Attribution Framework for Interpretability in Time Series Data.

With the rise in the employment of deep learning methods in safety-critical scenarios, interpretability is more essential than ever before. Although many different directions regarding interpretability have been explored for visual modalities, time series data has been neglected, with only a handful of methods tested due to their poor intelligibility. We approach the problem of interpretability in a novel way by proposing TSInsight, where we attach an auto-encoder to the classifier with a sparsity-inducing norm on its output and fine-tune it based on the gradients from the classifier and a reconstruction penalty. TSInsight learns to preserve features that are important for prediction by the classifier and suppresses those that are irrelevant, i.e., serves as a feature attribution method to boost the interpretability. In contrast to most other attribution frameworks, TSInsight is capable of generating both instance-based and model-based explanations. We evaluated TSInsight along with nine other commonly used attribution methods on eight different time series datasets to validate its efficacy. The evaluation results show that TSInsight naturally achieves output space contraction; therefore, it is an effective tool for the interpretability of deep time series models.

The privacy-explainability trade-off: unraveling the impacts of differential privacy and federated learning on attribution methods.

Since the advent of deep learning (DL), the field has witnessed a continuous stream of innovations. However, the translation of these advancements into practical applications has not kept pace, particularly in safety-critical domains where artificial intelligence (AI) must meet stringent regulatory and ethical standards. This is underscored by the ongoing research in eXplainable AI (XAI) and privacy-preserving machine learning (PPML), which seek to address some limitations associated with these opaque and data-intensive models. Despite brisk research activity in both fields, little attention has been paid to their interaction. This work is the first to thoroughly investigate the effects of privacy-preserving techniques on explanations generated by common XAI methods for DL models. A detailed experimental analysis is conducted to quantify the impact of private training on the explanations provided by DL models, applied to six image datasets and five time series datasets across various domains. The analysis comprises three privacy techniques, nine XAI methods, and seven model architectures. The findings suggest non-negligible changes in explanations through the implementation of privacy measures. Apart from reporting individual effects of PPML on XAI, the paper gives clear recommendations for the choice of techniques in real applications. By unveiling the interdependencies of these pivotal technologies, this research marks an initial step toward resolving the challenges that hinder the deployment of AI in safety-critical settings.

Challenges and perspectives in use of artificial intelligence to support treatment recommendations in clinical oncology.

Artificial intelligence (AI) promises to be the next revolutionary step in modern society. Yet, its role in all fields of industry and science need to be determined. One very promising field is represented by AI-based decision-making tools in clinical oncology leading to more comprehensive, personalized therapy approaches. In this review, the authors provide an overview on all relevant technical applications of AI in oncology, which are required to understand the future challenges and realistic perspectives for decision-making tools. In recent years, various applications of AI in medicine have been developed focusing on the analysis of radiological and pathological images. AI applications encompass large amounts of complex data supporting clinical decision-making and reducing errors by objectively quantifying all aspects of the data collected. In clinical oncology, almost all patients receive a treatment recommendation in a multidisciplinary cancer conference at the beginning and during their treatment periods. These highly complex decisions are based on a large amount of information (of the patients and of the various treatment options), which need to be analyzed and correctly classified in a short time. In this review, the authors describe the technical and medical requirements of AI to address these scientific challenges in a multidisciplinary manner. Major challenges in the use of AI in oncology and decision-making tools are data security, data representation, and explainability of AI-based outcome predictions, in particular for decision-making processes in multidisciplinary cancer conferences. Finally, limitations and potential solutions are described and compared for current and future research attempts.

[Descriptions of hospitalizations during the year 2019 among dialysis patients in France]. / Descriptions des hospitalisations au cours de l'année 2019 des patients dialysés en France.

INTRODUCTION: The possibility of indirect linkage between data from the French REIN registry and data from the National Health Data System makes it possible to enhance knowledge of the care pathway of patients with severe kidney disease. Taking an interest in hospitalizations makes it possible to understand the burden of this disease, both for patients, in terms of comorbidities, complications and quality of life but also for society in terms of care load, need for organization of the care offer and costs. METHODS: Patients on dialysis in 2019 were identified from REIN. The information on hospital stays comes from an extraction of the National Health Data System linked with REIN. The hospitalization rate corresponds to the number of hospital stays over the period, relative to the number of people at risk on dialysis over the period. RESULTS: After excluding dialysis sessions, 36,962 patients had a hospital stay in 2019: 36,738 in medicine, surgery, obstetrics, 2863 in rehabilitation, 531 in home hospitalizations and 201 in psychiatry facilities. This represents a total of 146,743 stays, including 140,372 in medicine, surgery, obstetrics. The hospitalization rate in medicine, surgery, obstetrics is 259 stays for 100 person-years. The reasons vary according to age, type of hospitalization (with or without overnight stays) and mode of discharge (home, death or transfer). The median amounts for a stay vary from €7920 in rehabilitation to €719 in medicine, surgery, obstetrics without overnight stay. Due to the volumes, stays for cardiovascular pathology represent a high total amount over the year. Hospitalizations for infection have significant median costs. DISCUSSION AND CONCLUSION: The results presented in this article confirm the important place of hospitalization in the trajectory of dialysis patients. Hospitalizations in medicine, surgery, obstetrics represent 96% of hospitalizations. For the first time, this study shows the very low place of hospitalizations in psychiatry and rehabilitation facilities.

Impact and benefit-cost ratio of a program for the management of latent tuberculosis infection among refugees in a region of Canada.

INTRODUCTION: The identification and treatment of latent tuberculosis infection (LTBI) among immigrants from high-incidence regions who move to low-incidence countries is generally considered an ineffective strategy because only ≈14% of them comply with the multiple steps of the 'cascade of care' and complete treatment. In the Estrie region of Canada, a refugee clinic was opened in 2009. One of its goals is LTBI management. METHODS: Key components of this intervention included: close collaboration with community organizations, integration within a comprehensive package of medical care for the whole family, timely delivery following arrival, shorter treatment through preferential use of rifampin, and risk-based selection of patients to be treated. Between 2009-2020, 5131 refugees were evaluated. To determine the efficacy and benefit-cost ratio of this intervention, records of refugees seen in 2010-14 (n = 1906) and 2018-19 (n = 1638) were reviewed. Cases of tuberculosis (TB) among our foreign-born population occurring before (1997-2008) and after (2009-2020) setting up the clinic were identified. All costs associated with TB or LTBI were measured. RESULTS: Out of 441 patients offered LTBI treatment, 374 (85%) were compliant. Adding other losses, overall compliance was 69%. To prevent one case of TB, 95.1 individuals had to be screened and 11.9 treated, at a cost of $16,056. After discounting, each case of TB averted represented $32,631, for a benefit-cost ratio of 2.03. Among nationals of the 20 countries where refugees came from, incidence of TB decreased from 68.2 (1997-2008) to 26.3 per 100,000 person-years (2009-2020). Incidence among foreign-born persons from all other countries not targeted by the intervention did not change. CONCLUSIONS: Among refugees settling in our region, 69% completed the LTBI cascade of care, leading to a 61% reduction in TB incidence. This intervention was cost-beneficial. Current defeatism towards LTBI management among immigrants and refugees is misguided. Compliance can be enhanced through simple measures.

The mechanisms underlying MMR deficiency in immunodeficiency-related non-Hodgkin lymphomas are different from those in other sporadic microsatellite instable neoplasms.

The spectrum of tumors showing microsatellite instability (MSI) has recently been enlarged to sporadic neoplasms whose incidence is favored in the context of chronic immunosuppression. We investigated the biological, therapeutic and clinical features associated with MSI in immunodeficiency-related non-Hodgkin lymphomas (ID-RL). MSI screening was performed in 275 ID-RL. MSI ID-RL were further analyzed for MMR gene expression and for BRAF/KRAS mutations since these genes are frequently altered in MSI cancers. We also assessed the expression of O(6)-methylguanine-DNA methyltransferase (MGMT), an enzyme whose inactivation has been reported in lymphomas and may help in the selection of MMR deficient clones. Unlike other sporadic MSI neoplasms, MSI ID-RL (N = 17) presented with heterogeneous MMR defects and no MLH1 promoter methylation. About one third of these tumors presented with normal expression of MLH1, MSH2, MSH6 and PMS2. They accumulated BRAF activating mutations (33%). Unlike other ID-RL, MSI ID-RL were primarily EBV-negative NHL of T-cell origin, and arose after long-term immunosuppression in patients who received azathioprine as part of their immunosuppressive regimen (p = 0.05) and/or who exhibited methylation-induced loss of expression of MGMT in tumor cells (p= 0.02). Overall, these results highlight that, in the context of deficient immune status, some MSI neoplasms arise through alternative mechanism when compared to other sporadic MSI neoplasms. They give the exact way how to make the diagnosis of MSI in these tumors and may help to define biological and clinicalrisk factors associated with their emergence in such a clinicalcontext.

Detoxication enzyme inducers modify cytokine production in rat mixed glial cells.

Pro-inflammatory cytokines, e.g. interleukin-1beta (IL-1beta), interleukin-6 (IL-6), tumor necrosis factor-alpha (TNFalpha) as well as neurotoxic molecules such as nitric oxide (NO), that are produced and released by activated glial cells, play an important role in inflammation and oxidative stress occurring during Multiple Sclerosis (MS). Reduction of these processes could therefore be of therapeutic interest. Dimethylfumarate (DMF) and sulforaphane (SP) are well known for their detoxicating properties. Furthermore, they have anti-inflammatory effects as shown clinically by the treatment of inflammatory skin diseases. However, their detoxication and anti-inflammatory action on brain-derived cells is unknown. In the present study we have studied, within the same concentration range, the anti-inflammatory and detoxicating effects of DMF and SP on the production and release of mediators of inflammation and detoxication from lipopolysaccharide (LPS) activated primary co-cultures of rat microglial and astroglial cells. DMF and SP attenuated the LPS-induced production and release of TNFalpha, IL-1beta, IL-6 and NO. In addition, DMF and SP increase both mRNA level and activity of NAD(P)H:quinone reductase (NQO-1), a detoxication enzyme, as well as the cellular glutathione content. We conclude that DMF or SP simultaneously can (1) reduce mediators of inflammation and (2) enhance detoxication enzymes in LPS stimulated co-cultures of astroglial and microglial cells. This double-sided effect could potentially be of therapeutic interest.

Nonsense-mediated mRNA decay impacts MSI-driven carcinogenesis and anti-tumor immunity in colorectal cancers.

Nonsense-mediated mRNA Decay (NMD) degrades mutant mRNAs containing premature termination codon (PTC-mRNAs). Here we evaluate the consequence of NMD activity in colorectal cancers (CRCs) showing microsatellite instability (MSI) whose progression is associated with the accumulation of PTC-mRNAs encoding immunogenic proteins due to frameshift mutations in coding repeat sequences. Inhibition of UPF1, one of the major NMD factors, was achieved by siRNA in the HCT116 MSI CRC cell line and the resulting changes in gene expression were studied using expression microarrays. The impact of NMD activity was also investigated in primary MSI CRCs by quantifying the expression of several mRNAs relative to their mutational status and to endogenous UPF1 and UPF2 expression. Host immunity developed against MSI cancer cells was appreciated by quantifying the number of CD3epsilon-positive tumor-infiltrating lymphocytes (TILs). UPF1 silencing led to the up-regulation of 1251 genes in HCT116, among which a proportion of them (i.e. 38%) significantly higher than expected by chance contained a coding microsatellite (P<2x10(-16)). In MSI primary CRCs, UPF1 was significantly over-expressed compared to normal adjacent mucosa (P<0.002). Our data provided evidence for differential decay of PTC-mRNAs compared to wild-type that was positively correlated to UPF1 endogenous expression level (P = 0.02). A negative effect of UPF1 and UPF2 expression on the host's anti-tumor response was observed (P<0.01). Overall, our results show that NMD deeply influences MSI-driven tumorigenesis at the molecular level and indicate a functional negative impact of this system on anti-tumor immunity whose intensity has been recurrently shown to be an independent factor of favorable outcome in CRCs.

[Clinical and molecular consequences of microsatellite instability in human cancers]. / Conséquences cliniques et moléculaires de l'instabilité des microsatellites dans les cancers humains.

During each cell division, DNA polymerase makes mistakes while copying DNA. These errors, more frequent at the level of repeated sequences called microsatellites are normally repaired by a system called MMR (mismatch repair). Tumors defective in their MMR system accumulate mutations (deletions and insertions of some nucleotides) at the level of microsatellites and are called MSI (microsatellite instability). Microsatellites are numerous and scattered throughout the genome, in coding and non-coding regions. The instability of non-coding microsatellites is not known to have a major role in the process of cell transformation, but is a good indicator of the MSI status. On the other hand, instability by deletion or insertion in a coding region leads to a frameshift within the gene containing the repeat. The consequence is, the more often, the inactivation of this gene that potentially plays a role in initiation and/or MSI tumor progression. The MSI phenotype was first described in about 15 % of colorectal cancers that maybe of sporadic or hereditary (Lynch syndrome, or HNPCC for hereditary non-polyposis colorectal cancer) origin. It is also associated with about 15 % of gastric and endometrial tumors, and to a lesser extent with other human tumors. Besides a fundamental interest because of its original transformation mechanism, the analysis of MSI tumors is also important for clinical reasons. It was indeed shown that MSI tumors were associated with a better prognosis than non-MSI (also called MSS for microsatellite stable) tumors, and responded differently to conventional chemotherapeutic drugs used for the management of colorectal cancers. All these points will be discussed in details in the present review.

Regulation of alpha1,3galactosyltransferase expression in pig endothelial cells. Implications for xenotransplantation.

The disaccharide galactose(alpha)1,3 galactose (the alphaGal epitope) is the major xenoantigen responsible for the hyperacute vascular rejection occurring in pig-to-primates organ transplantation. The synthesis of the alphaGal epitope is catalyzed by the enzyme alpha1,3-galactosyltransferase (alpha1,3GalT). To be able to control porcine alpha1,3GalT gene expression specifically, we have analyzed the upstream portion of the alpha1,3GalT gene, and identified the regulatory sequences. Porcine alpha1,3GalT transcripts were detected by 5' RACE analysis, and the corresponding genomic sequences were isolated from a phage library. The porcine alpha1,3GalT gene consists of at least 10 different exons, four of which contain 5' untranslated sequence. Four distinct promoters, termed A-D, drive alpha1,3GalT gene transcription in porcine cells. A combination of alternative promoter usage and alternative splicing produces a series of transcripts that differ in their 5' portion, but encode the same protein. Promoters A-C have been isolated, and functionally characterized using luciferase reporter gene assays in transfected porcine endothelial cells (PEC-A). Promoter preference in porcine endothelial cells was estimated on the basis of relative transcript levels as determined by real-time quantitative PCR. More than 90% of the alpha1,3GalT transcripts in PEC-A cells originate from promoter B, which has characteristics of a housekeeping gene promoter. While promoter preference remains unchanged, alpha1,3GalT mRNA levels increase by 50% in 12 h upon tumour necrosis factor alpha-activation of PEC-A cells. However, the magnitude of this change induced by inflammatory conditions could be insufficient to affect cell surface alpha1,3-galactosylation.