Serum Neurofilament Light Chain in Replication Factor Complex Subunit 1 CANVAS and Disease Spectrum

Background Biallelic intronic AAGGG repeat expansions in the replication factor complex subunit 1 ( RFC1 ) gene were identified as the leading cause of cerebellar ataxia, neuropathy, vestibular areflexia syndrome. Patients exhibit significant clinical heterogeneity and variable disease course, but no potential biomarker has been identified to date. Objectives In this multicenter cross‐sectional study, we aimed to evaluate neurofilament light (NfL) chain serum levels in a cohort of RFC1 disease patients and to correlate NfL serum concentrations with clinical phenotype and disease severity. Methods Sixty‐one patients with genetically confirmed RFC1 disease and 48 healthy controls (HCs) were enrolled from six neurological centers. Serum NfL concentration was measured using the single molecule array assay technique. Results Serum NfL concentration was significantly higher in patients with RFC1 disease compared to age‐ and‐sex‐matched HCs ( P  < 0.0001). NfL level showed a moderate correlation with age in both HCs ( r  = 0.4353, P  = 0.0020) and patients ( r  = 0.4092, P  = 0.0011). Mean NfL concentration appeared to be significantly higher in patients with cerebellar involvement compared to patients without cerebellar dysfunction (27.88 vs. 21.84 pg/mL, P  = 0.0081). The association between cerebellar involvement and NfL remained significant after controlling for age and sex (β = 0.260, P  = 0.034). Conclusions Serum NfL levels are significantly higher in patients with RFC1 disease compared to HCs and correlate with cerebellar involvement. Longitudinal studies are warranted to assess its change over time.

Uncertainty quantification in multi‐class segmentation: Comparison between Bayesian and non‐Bayesian approaches in a clinical perspective

Abstract Background Automatic segmentation techniques based on Convolutional Neural Networks (CNNs) are widely adopted to automatically identify any structure of interest from a medical image, as they are not time consuming and not subject to high intra‐ and inter‐operator variability. However, the adoption of these approaches in clinical practice is slowed down by some factors, such as the difficulty in providing an accurate quantification of their uncertainty. Purpose This work aims to evaluate the uncertainty quantification provided by two Bayesian and two non‐Bayesian approaches for a multi‐class segmentation problem, and to compare the risk propensity among these approaches, considering CT images of patients affected by renal cancer (RC). Methods Four uncertainty quantification approaches were implemented in this work, based on a benchmark CNN currently employed in medical image segmentation: two Bayesian CNNs with different regularizations (Dropout and DropConnect), named BDR and BDC, an ensemble method (Ens) and a test‐time augmentation (TTA) method. They were compared in terms of segmentation accuracy, using the Dice score, uncertainty quantification, using the ratio of correct‐certain pixels (RCC) and incorrect‐uncertain pixels (RIU), and with respect to inter‐observer variability in manual segmentation. They were trained with the Kidney and Kidney Tumor Segmentation Challenge launched in 2021 (Kits21), for which multi‐class segmentations of kidney, RC, and cyst on 300 CT volumes are available. Moreover, they were tested considering this and other two public renal CT datasets. Results Accuracy results achieved large differences across the structures of interest for all approaches, with an average Dice score of 0.92, 0.58, and 0.21 for kidney, tumor, and cyst, respectively. In terms of uncertainties, TTA provided the highest uncertainty, followed by Ens and BDC, whereas BDR provided the lowest, and minimized the number of incorrect certain pixels worse than the other approaches. Again, large differences were seen across the three structures in terms of RCC and RIU. These metrics were associated with different risk propensity, as BDR was the most risk‐taking approach, able to provide higher accuracy in its prediction, but failing to assign uncertainty on incorrect segmentation in every case. The other three approaches were more conservative, providing large uncertainty regions, with the drawback of giving alert also on correct areas. Finally, the analysis of the inter‐observer segmentation variability showed a significant variation among the four approaches on the external dataset, with BDR reporting the lowest agreement (Dice = 0.82), and TTA obtaining the highest score (Dice = 0.94). Conclusions Our outcomes highlight the importance of quantifying the segmentation uncertainty and that decision‐makers can choose the approach most in line with the risk propensity degree required by the application and their policy.

Indirect 1 H–[ 13 C] MRS of the human brain at 7 T using a 13 C‐birdcage coil and eight transmit–receive 1 H‐dipole antennas with a 32‐channel 1 H‐receive array

The neuronal tricarboxylic acid and glutamate/glutamine (Glu/Gln) cycles play important roles in brain function. These processes can be measured in vivo using dynamic 1 H–[ 13 C] MRS during administration of 13 C‐labeled glucose. Proton‐observed carbon‐edited (POCE) MRS enhances the signal‐to‐noise ratio (SNR) compared with direct 13 C‐MRS. Ultra‐high field further boosts the SNR and increases spectral dispersion; however, even at 7 T, Glu and Gln 1 H‐resonances may overlap. Further gain can be obtained with selective POCE (selPOCE). Our aim was to create a setup for indirect dynamic 1 H–[ 13 C] MRS in the human brain at 7 T. A home‐built non‐shielded transmit–receive 13 C‐birdcage head coil with eight transmit–receive 1 H‐dipole antennas was used together with a 32‐channel 1 H‐receive array. Electromagnetic simulations were carried out to ensure that acquisitions remained within local and global head SAR limits. POCE‐MRS was performed using slice‐selective excitation with semi‐localization by adiabatic selective refocusing (sLASER) and stimulated echo acquisition mode (STEAM) localization, and selPOCE‐MRS using STEAM. Sequences were tested in a phantom containing non‐enriched Glu and Gln, and in three healthy volunteers during uniformly labeled 13 C‐glucose infusions. In one subject the voxel position was alternated between bi‐frontal and bi‐occipital placement within one session. [4‐ 13 C]Glu‐H4 and [4‐ 13 C]Gln‐H4 signals could be separately detected using both STEAM‐POCE and STEAM‐selPOCE in the phantom. In vivo, [4,5‐ 13 C]Glx could be detected using both sLASER‐POCE and STEAM‐POCE, with similar sensitivities, but [4,5‐ 13 C]Glu and [4,5‐ 13 C]Gln signals could not be completely resolved. STEAM‐POCE was alternately performed bi‐frontal and bi‐occipital within a single session without repositioning of the subject, yielding similar results. With STEAM‐selPOCE, [4,5‐ 13 C]Glu and [4,5‐ 13 C]Gln could be clearly separated. We have shown that with our setup indirect dynamic 1 H–[ 13 C] MRS at 7 T is feasible in different locations in the brain within one session, and by using STEAM‐selPOCE it is possible to separate Glu from Gln in vivo while obtaining high quality spectra.

Hepatic and adipose tissue transcriptome analysis highlights a commonly deregulated autophagic pathway in severe MASLD

Abstract Objective The incidence of metabolic dysfunction‐associated steatotic liver disease (MASLD) is rapidly ramping up due to the spread of obesity, which is characterized by expanded and dysfunctional visceral adipose tissue (VAT). Previous studies have investigated the hepatic transcriptome across MASLD, whereas few studies have focused on VAT. Methods We performed RNA sequencing in 167 hepatic samples from patients with obesity and in a subset of 79 matched VAT samples. Circulating cathepsin D (CTSD), a lysosomal protease, was measured by ELISA, whereas the autophagy‐lysosomal pathway was assessed by Western blot in hepatic and VAT samples ( n  = 20). Results Inflammation, extracellular matrix remodeling, and mitochondrial dysfunction were upregulated in severe MASLD in both tissues, whereas autophagy and oxidative phosphorylation were reduced. Tissue comparative analysis revealed 13 deregulated genes, including CTSD, which showed the most robust diagnostic accuracy in discriminating mild and severe MASLD. CTSD expression correlated with circulating protein, whose increase was further validated in 432 histologically characterized MASLD patients, showing a high accuracy in foreseeing severe liver injury. In addition, the assessment of serum CTSD increased the performance of fibrosis 4 in diagnosing advanced disease. Conclusions By comparing the hepatic and VAT transcriptome during MASLD, we refined the concept by which CTSD may represent a potential biomarker of severe disease.

To retrieve values of albuminuria to define the size of CKD patients eligible to SGLT‐2Is : An explorative analysis using a primary care database

Purpose To address missingness of albuminuria values, which establish the eligibility to SGLT‐2Is for patients with CKD, using the multiple imputation (MI) method. Methods We selected patients aged 18 or older and diagnosed with CKD in a primary care database. Those with severe CKD and/or previously treated with SGLT‐2Is were excluded. Then, we collected all available information on albuminuria within 90 days the measure of GFR. A value of albumin‐creatinine ratio (ACR) ranging 200–5000 mg/g or otherwise was the response variable on which we ran MI. Using logistic regression, odds ratios (OR) and related 95% confidence intervals (CI) were estimated for each covariate toward the response variable for both full and imputed dataset. Results The determinants showed consistent estimates between the full and imputed datasets as demonstrated by the overlaps of the CIs and the similar point estimates. As expected, there were some exceptions, such as diabetes (OR of 1.2 vs. 0.5) and use of diabetic medications (OR of 1.0 vs. 2.1) and/or statins (OR of 1.1 vs. 1.8). Conclusions Besides being a reminder for GPs to prescribe and register albuminuria in certain patients' categories, these determinants might be translated into an operational algorithm to input ACR values in administrative data sources. Scenarios for the reimbursement criteria regulating SGLT‐2Is to treat CKD would be therefore simulated on more inferable estimates.

Machine Learning based Noise Characterization and Correction on Neutral Atoms NISQ Devices

Neutral atoms devices represent a promising technology using optical tweezers to geometrically arrange atoms and modulated laser pulses to control their quantum states. They are exploited as noisy intermediate‐scale quantum (NISQ) processors. Indeed, like all real quantum devices, they are affected by noise introducing errors in the computation. Therefore, it is important to understand and characterize the noise sources and possibly to correct them. Here, two machine‐learning based approaches are proposed respectively to estimate the noise parameters and to mitigate their effects using only measurements of the final quantum state. Our analysis is then tested on a real neutral atom platform, comparing our predictions with a priori estimated parameters. It turns out that increasing the number of atoms is less effective than using more measurements on a smaller scale. The agreement is not always good but this may be due to the limited amount of real data that are obtained from a still under development device. Finally, reinforcement learning is employed to design a pulse that mitigates the noise effects. Our machine learning‐based approach is espected to be very useful for the noise benchmarking of NISQ processors and, more in general, of real quantum technologies.

Extranodal natural killer/T cell lymphoma of the skeletal muscle

Lutetium [177Lu]-DOTA-TATE in gastroenteropancreatic-neuroendocrine tumours: rationale, design and baseline characteristics of the Italian prospective observational (REAL-LU) study

The surgical learning curve for percutaneous Zadek osteotomy for treatment of insertional achilles tendinopathy

An AI-based approach driven by genotypes and phenotypes to uplift the diagnostic yield of genetic diseases