Advances in the Use of Deep Learning for the Analysis of Magnetic Resonance Image in Neuro-Oncology.

Machine Learning is entering a phase of maturity, but its medical applications still lag behind in terms of practical use. The field of oncological radiology (and neuro-oncology in particular) is at the forefront of these developments, now boosted by the success of Deep-Learning methods for the analysis of medical images. This paper reviews in detail some of the most recent advances in the use of Deep Learning in this field, from the broader topic of the development of Machine-Learning-based analytical pipelines to specific instantiations of the use of Deep Learning in neuro-oncology; the latter including its use in the groundbreaking field of ultra-low field magnetic resonance imaging.

AI-Based Glioma Grading for a Trustworthy Diagnosis: An Analytical Pipeline for Improved Reliability.

Glioma is the most common type of tumor in humans originating in the brain. According to the World Health Organization, gliomas can be graded on a four-stage scale, ranging from the most benign to the most malignant. The grading of these tumors from image information is a far from trivial task for radiologists and one in which they could be assisted by machine-learning-based decision support. However, the machine learning analytical pipeline is also fraught with perils stemming from different sources, such as inadvertent data leakage, adequacy of 2D image sampling, or classifier assessment biases. In this paper, we analyze a glioma database sourced from multiple datasets using a simple classifier, aiming to obtain a reliable tumor grading and, on the way, we provide a few guidelines to ensure such reliability. Our results reveal that by focusing on the tumor region of interest and using data augmentation techniques we significantly enhanced the accuracy and confidence in tumor classifications. Evaluation on an independent test set resulted in an AUC-ROC of 0.932 in the discrimination of low-grade gliomas from high-grade gliomas, and an AUC-ROC of 0.893 in the classification of grades 2, 3, and 4. The study also highlights the importance of providing, beyond generic classification performance, measures of how reliable and trustworthy the model's output is, thus assessing the model's certainty and robustness.

Rebleeding and mortality risk are increased by ACLF but reduced by pre-emptive TIPS.

BACKGROUND & AIMS: The relationship between acute-on-chronic liver failure (ACLF) and acute variceal bleeding (AVB) is poorly understood. Specifically, the prevalence and prognosis of ACLF in the context of AVB is unclear, while the role of transjugular intrahepatic portosystemic shunt (TIPS) in the management in patients with ACLF has not been described to date. METHODS: A multicenter, international, observational study was conducted in 2,138 patients from 34 centers between 2011 and 2015. ACLF was defined and graded according to the EASL-CLIF consortium definition. Placement of pre-emptive TIPS (pTIPS) was based on individual center policy. Patients were followed-up for 1 year, until death or liver transplantation. Cox regression and competing risk models (Gray's test) were used to identify independent predictors of rebleeding or mortality. RESULTS: At admission, 380/2,138 (17.8%) patients had ACLF according to EASL-CLIF criteria (grade 1: 38.7%; grade 2: 39.2%; grade 3: 22.1%). The 42-day rebleeding (19% vs. 10%; p <0.001) and mortality (47% vs. 10%; p <0.001) rates were higher in patients with ACLF and increased with ACLF grades. Of note, the presence of ACLF was independently associated with rebleeding and mortality. pTIPS placement improved survival in patients with ACLF at 42 days and 1 year. This effect was also observed in propensity score matching analysis of 66 patients with ACLF, of whom 44 received pTIPs and 22 did not. CONCLUSIONS: This large multicenter international real-life study identified ACLF at admission as an independent predictor of rebleeding and mortality in patients with AVB. Moreover, pTIPS was associated with improved survival in patients with ACLF and AVB. LAY SUMMARY: Acute variceal bleeding is a deadly complication of liver cirrhosis that results from severe portal hypertension. This study demonstrates that the presence of acute-on-chronic liver failure (ACLF) is the strongest predictor of mortality in patients with acute variceal bleeding. Importantly, patients with ACLF and acute variceal (re)bleeding benefit from pre-emptive (early) placement of a transjugular intrahepatic portosystemic shunt.

HDL-related biomarkers are robust predictors of survival in patients with chronic liver failure.

BACKGROUND & AIMS: High-density lipoprotein cholesterol (HDL-C) levels are reduced in patients with chronic liver disease and inversely correlate with disease severity. During acute conditions such as sepsis, HDL-C levels decrease rapidly and HDL particles undergo profound changes in their composition and function. We aimed to determine whether indices of HDL quantity and quality associate with progression and survival in patients with advanced liver disease. METHODS: HDL-related biomarkers were studied in 508 patients with compensated or decompensated cirrhosis (including acute-on-chronic liver failure [ACLF]) and 40 age- and gender-matched controls. Specifically, we studied levels of HDL-C, its subclasses HDL2-C and HDL3-C, and apolipoprotein A1 (apoA-I), as well as HDL cholesterol efflux capacity as a metric of HDL functionality. RESULTS: Baseline levels of HDL-C and apoA-I were significantly lower in patients with stable cirrhosis compared to controls and were further decreased in patients with acute decompensation (AD) and ACLF. In stable cirrhosis (n = 228), both HDL-C and apoA-I predicted the development of liver-related complications independently of model for end-stage liver disease (MELD) score. In patients with AD, with or without ACLF (n = 280), both HDL-C and apoA-I were MELD-independent predictors of 90-day mortality. On ROC analysis, both HDL-C and apoA-I had high diagnostic accuracy for 90-day mortality in patients with AD (AUROCs of 0.79 and 0.80, respectively, similar to that of MELD 0.81). On Kaplan-Meier analysis, HDL-C <17 mg/dl and apoA-I <50 mg/dl indicated poor short-term survival. The prognostic accuracy of HDL-C was validated in a large external validation cohort of 985 patients with portal hypertension due to advanced chronic liver disease (AUROCs HDL-C: 0.81 vs. MELD: 0.77). CONCLUSION: HDL-related biomarkers are robust predictors of disease progression and survival in chronic liver failure. LAY SUMMARY: People who suffer from cirrhosis (scarring of the liver) have low levels of cholesterol carried by high-density lipoproteins (HDL-C). These alterations are connected to inflammation, which is a problem in severe liver disease. Herein, we show that reduced levels of HDL-C and apolipoprotein A-I (apoA-I, the main protein carried by HDL) are closely linked to the severity of liver failure, its complications and survival. Both HDL-C and apoA-I can be easily measured in clinical laboratories and are as good as currently used prognostic scores calculated from several laboratory values by complex formulas.

Addressing Profiles of Systemic Inflammation Across the Different Clinical Phenotypes of Acutely Decompensated Cirrhosis.

Background: Patients with acutely decompensated cirrhosis (AD) may or may not develop acute-on-chronic liver failure (ACLF). ACLF is characterized by high-grade systemic inflammation, organ failures (OF) and high short-term mortality. Although patients with AD cirrhosis exhibit distinct clinical phenotypes at baseline, they have low short-term mortality, unless ACLF develops during follow-up. Because little is known about the association of profile of systemic inflammation with clinical phenotypes of patients with AD cirrhosis, we aimed to investigate a battery of markers of systemic inflammation in these patients. Methods: Upon hospital admission baseline plasma levels of 15 markers (cytokines, chemokines, and oxidized albumin) were measured in 40 healthy controls, 39 compensated cirrhosis, 342 AD cirrhosis, and 161 ACLF. According to EASL-CLIF criteria, AD cirrhosis was divided into three distinct clinical phenotypes (AD-1: Creatinine<1.5, no HE, no OF; AD-2: creatinine 1.5-2, and or HE grade I/II, no OF; AD-3: Creatinine<1.5, no HE, non-renal OF). Results: Most markers were slightly abnormal in compensated cirrhosis, but markedly increased in AD. Patients with ACLF exhibited the largest number of abnormal markers, indicating "full-blown" systemic inflammation (all markers). AD-patients exhibited distinct systemic inflammation profiles across three different clinical phenotypes. In each phenotype, activation of systemic inflammation was only partial (30% of the markers). Mortality related to each clinical AD-phenotype was significantly lower than mortality associated with ACLF (p < 0.0001 by gray test). Among AD-patients baseline systemic inflammation (especially IL-8, IL-6, IL-1ra, HNA2 independently associated) was more intense in those who had poor 28-day outcomes (ACLF, death) than those who did not experience these outcomes. Conclusions: Although AD-patients exhibit distinct profiles of systemic inflammation depending on their clinical phenotypes, all these patients have only partial activation of systemic inflammation. However, those with the most extended baseline systemic inflammation had the highest the risk of ACLF development and death.