Beta-defensin index: A functional biomarker for oral cancer detection.

There is an unmet clinical need for a non-invasive and cost-effective test for oral squamous cell carcinoma (OSCC) that informs clinicians when a biopsy is warranted. Human beta-defensin 3 (hBD-3), an epithelial cell-derived anti-microbial peptide, is pro-tumorigenic and overexpressed in early-stage OSCC compared to hBD-2. We validate this expression dichotomy in carcinoma in situ and OSCC lesions using immunofluorescence microscopy and flow cytometry. The proportion of hBD-3/hBD-2 levels in non-invasively collected lesional cells compared to contralateral normal cells, obtained by ELISA, generates the beta-defensin index (BDI). Proof-of-principle and blinded discovery studies demonstrate that BDI discriminates OSCC from benign lesions. A multi-center validation study shows sensitivity and specificity values of 98.2% (95% confidence interval [CI] 90.3-99.9) and 82.6% (95% CI 68.6-92.2), respectively. A proof-of-principle study shows that BDI is adaptable to a point-of-care assay using microfluidics. We propose that BDI may fulfill a major unmet need in low-socioeconomic countries where pathology services are lacking.

Intrabiliary metastasis of colorectal mucinous adenocarcinoma mimicking choledocholithiasis 18 years after the primary tumor.

This case report presents a 62-year-old male who had previously undergone curative colectomy and neoadjuvant chemotherapy in 2005 for colorectal cancer. He presented with jaundice, which was initially attributed to choledocholithiasis. After cholecystectomy and repeat ERCPs, hyperbilirubinemia persisted. There was persistent dilation of the right posterior duct on imaging, concerning for biliary stricture, possibly due to cholangiocarcinoma or intraductal papillary neoplasm. During a right posterior hepatectomy, a peripheral liver lesion was found in association with the dilated bile duct. On frozen evaluation, the lesion was found to be invasive adenocarcinoma. The final pathology was compatible with a metastatic mucinous adenocarcinoma of colonic origin. A repeat colonoscopy was done with no recurrence or new lesion in the colon. This case underscores the challenges associated with diagnosing biliary issues and assessing liver lesions in patients with a remote history of cancer. It raises the question of when and whether, after primary cancer treatment, it becomes safe to explore alternative diagnoses without immediately suspecting metastasis. Another significant challenge arises in ascertaining the most suitable therapeutic approaches for these patients. This is because these extremely late recurrences might be linked to an indolent, slow-growing type of tumor, but also have been linked to cancer stem cells, and as any recurrence, demands attention.

Ab initio quantum chemistry with neural-network wavefunctions.

Deep learning methods outperform human capabilities in pattern recognition and data processing problems and now have an increasingly important role in scientific discovery. A key application of machine learning in molecular science is to learn potential energy surfaces or force fields from ab initio solutions of the electronic Schrödinger equation using data sets obtained with density functional theory, coupled cluster or other quantum chemistry (QC) methods. In this Review, we discuss a complementary approach using machine learning to aid the direct solution of QC problems from first principles. Specifically, we focus on quantum Monte Carlo methods that use neural-network ansatzes to solve the electronic Schrödinger equation, in first and second quantization, computing ground and excited states and generalizing over multiple nuclear configurations. Although still at their infancy, these methods can already generate virtually exact solutions of the electronic Schrödinger equation for small systems and rival advanced conventional QC methods for systems with up to a few dozen electrons.

Discovering Quantum Phase Transitions with Fermionic Neural Networks.

Deep neural networks have been very successful as highly accurate wave function Ansätze for variational Monte Carlo calculations of molecular ground states. We present an extension of one such Ansatz, FermiNet, to calculations of the ground states of periodic Hamiltonians, and study the homogeneous electron gas. FermiNet calculations of the ground-state energies of small electron gas systems are in excellent agreement with previous initiator full configuration interaction quantum Monte Carlo and diffusion Monte Carlo calculations. We investigate the spin-polarized homogeneous electron gas and demonstrate that the same neural network architecture is capable of accurately representing both the delocalized Fermi liquid state and the localized Wigner crystal state. The network converges on the translationally invariant ground state at high density and spontaneously breaks the symmetry to produce the crystalline ground state at low density, despite being given no a priori knowledge that a phase transition exists.

Response to Comment on "Pushing the frontiers of density functionals by solving the fractional electron problem".

Gerasimov et al. claim that the ability of DM21 to respect fractional charge (FC) and fractional spin (FS) conditions outside of the training set has not been demonstrated in our paper. This is based on (i) asserting that the training set has a ~50% overlap with our bond-breaking benchmark (BBB) and (ii) questioning the validity and accuracy of our other generalization examples. We disagree with their analysis and believe that the points raised are either incorrect or not relevant to the main conclusions of the paper and to the assessment of general quality of DM21.

Magnetic control of tokamak plasmas through deep reinforcement learning.

Nuclear fusion using magnetic confinement, in particular in the tokamak configuration, is a promising path towards sustainable energy. A core challenge is to shape and maintain a high-temperature plasma within the tokamak vessel. This requires high-dimensional, high-frequency, closed-loop control using magnetic actuator coils, further complicated by the diverse requirements across a wide range of plasma configurations. In this work, we introduce a previously undescribed architecture for tokamak magnetic controller design that autonomously learns to command the full set of control coils. This architecture meets control objectives specified at a high level, at the same time satisfying physical and operational constraints. This approach has unprecedented flexibility and generality in problem specification and yields a notable reduction in design effort to produce new plasma configurations. We successfully produce and control a diverse set of plasma configurations on the Tokamak à Configuration Variable1,2, including elongated, conventional shapes, as well as advanced configurations, such as negative triangularity and 'snowflake' configurations. Our approach achieves accurate tracking of the location, current and shape for these configurations. We also demonstrate sustained 'droplets' on TCV, in which two separate plasmas are maintained simultaneously within the vessel. This represents a notable advance for tokamak feedback control, showing the potential of reinforcement learning to accelerate research in the fusion domain, and is one of the most challenging real-world systems to which reinforcement learning has been applied.

Pushing the frontiers of density functionals by solving the fractional electron problem.

Density functional theory describes matter at the quantum level, but all popular approximations suffer from systematic errors that arise from the violation of mathematical properties of the exact functional. We overcame this fundamental limitation by training a neural network on molecular data and on fictitious systems with fractional charge and spin. The resulting functional, DM21 (DeepMind 21), correctly describes typical examples of artificial charge delocalization and strong correlation and performs better than traditional functionals on thorough benchmarks for main-group atoms and molecules. DM21 accurately models complex systems such as hydrogen chains, charged DNA base pairs, and diradical transition states. More crucially for the field, because our methodology relies on data and constraints, which are continually improving, it represents a viable pathway toward the exact universal functional.

Imaging features of gastrointestinal toxicity in non-small cell lung cancer patients treated with erlotinib: A single institute 13-year experience.

OBJECTIVES: To investigate the imaging features of erlotinib-associated gastrointestinal toxicity (GT) in patients with non-small cell lung cancer (NSCLC). MATERIALS AND METHODS: The electronic medical records of 157 patients with NSCLC who received erlotinib between 2005 and 2018 were retrospectively reviewed to identify patients with GT. Clinical and radiologic evidence of erlotinib-associated GT was evaluated. Imaging findings were cross-referenced with clinical presentation, management, and outcomes. RESULTS: 24 (15%) patients (16 women; median age, 68 years) with radiologic evidence of GT were identified. The median time to detection of GT on imaging was 4.5 months (range: 0-58 months). 3/24 (12.5%) patients had no clinical symptoms, but GT was radiologically identified. Erlotinib-associated GT manifested in the large bowel in either a diffuse (42%) or segmental (58%) pattern. The most common imaging finding was fluid-filled bowel (23/24, 96%). CONCLUSION: Erlotinib-associated GT was identified in 15% patients with NSCLC. Fluid-filled colon and segmental involvement were the most common imaging manifestations.

Decoding the Genomic Report for Radiologists.

OBJECTIVE. The purpose of this review is to provide a guide for radiologists that explains the language and format of modern genomic reports and summarizes the relevance of this information for modern oncologic imaging. CONCLUSION. Genomic testing plays a critical role in guiding oncologic therapies in the age of targeted treatments. Understanding and interpreting genomic reports is a valuable skill for radiologists involved with oncologic imaging interpretation.

Imaging and clinical manifestations of immune checkpoint inhibitor-related colitis in cancer patients treated with monotherapy or combination therapy.

PURPOSE: To determine the frequency, imaging, and clinical manifestations of immune checkpoint inhibitor (ICI)-related colitis in cancer patients on monotherapy or combination therapy. METHODS: The electronic medical records of 1044 cancer patients who received ICIs were retrospectively reviewed to identify 48 patients who had a clinical diagnosis of immune-related colitis. Imaging studies were reviewed to identify patients with imaging manifestations of colitis. Demographic data, type of ICIs, symptoms, presence of other immune-related adverse events (irAEs), and management were recorded. RESULTS: There was imaging evidence of immune-related colitis in 34 patients (24 men; median age: 63.5 years). The median time to onset of colitis was 75 days (IQR 25-75, 49.5-216 days) in patients receiving monotherapy (group 1) and 78 days (IQR 25-75, 44.3-99.5 days) in patients undergoing combination therapy (group 2) following start of ICI. Symptoms included diarrhea (91.1% [31 of 34]), nausea/vomiting (52.9% [18 of 34]), and abdominal pain (52.9% [18 of 34]). The most common imaging findings were bowel wall thickening (97% [33 of 34]) and fluid-filled colon (82.3% [28 of 34]). Colitis was diffuse in 21 of 34 (61.8%) patients. Imaging manifestations did not differ between the two groups (p > 0.05). Steroids and antibiotics were used to treat colitis in 29 of 34 (85.2%) and 13 of 34 (38.2%) patients, respectively. No patients in group 1 experienced concurrent irAEs, but 5 of 18 (27.8%) of patients in group 2 had other irAEs (p = 0.046). CONCLUSION: Immune-related colitis occurred in 3.3% of patients receiving ICIs with bowel wall thickening, fluid-filled colon and pancolitis being the most common imaging manifestations. Imaging manifestations did not differ between patients receiving monotherapy or combination therapy. However, concurrent irAEs were significantly observed in patients undergoing combination therapy.

Primary Mediastinal Large B-Cell Lymphoma: A Review for Radiologists.

OBJECTIVE. The purpose of this article is to provide a primer for radiologists focused on integrating the radiologic, pathologic, and clinical features of primary mediastinal large B-cell lymphoma (PMLBCL). CONCLUSION. PMLBCL is a unique subtype of lymphoma that poses diagnostic and therapeutic challenges to the fields of radiology and oncology. Knowledge of this distinctive clinical-pathologic entity and its associated imaging and clinical features is critical for radiologists.

Sentinel Node Biopsy in Young Patients with Atypical Melanocytic Tumors of the Head and Neck.

OBJECTIVE: To examine the diagnostic value of the sentinel lymph node biopsy in pediatric through young adult head and neck melanocytic tumors of unknown malignant potential. STUDY DESIGN: Retrospective case series. SETTING: Single academic institution. SUBJECTS AND METHODS: Demographics, histology, and outcomes were examined in 14 patients aged 4 to 24 years with head and neck melanocytic tumors of unknown malignant potential. Information on age at diagnosis, primary lesion characteristics, and sentinel lymph node biopsy were compared. RESULTS: Of 14 patients meeting criteria for head and neck melanocytic tumors of unknown malignant potential, 8 patients underwent sentinel lymph node biopsy (57%). Of those, 4 biopsies (50%) had positive sentinel nodes. All patients undergoing sentinel lymph node biopsy had primary lesions greater than 1 mm depth or mitotic rate of at least 1 mitosis per mm2. No patients had recurrence of their primary lesion at time of follow-up. CONCLUSION: Our data show a high rate of node-positive sentinel lymph node biopsy for pediatric and young adult head and neck patients with melanocytic tumors of unknown malignant potential, supporting the value of sentinel lymph node biopsy in this population.

Non-classical monocytes are biased progenitors of wound healing macrophages during soft tissue injury.

Successful tissue repair requires the activities of myeloid cells such as monocytes and macrophages that guide the progression of inflammation and healing outcome. Immunoregenerative materials leverage the function of endogenous immune cells to orchestrate complex mechanisms of repair; however, a deeper understanding of innate immune cell function in inflamed tissues and their subsequent interactions with implanted materials is necessary to guide the design of these materials. Blood monocytes exist in two primary subpopulations, characterized as classical inflammatory or non-classical. While classical monocytes extravasate into inflamed tissue and give rise to macrophages or dendritic cells, the recruitment kinetics and functional role of non-classical monocytes remains unclear. Here, we demonstrate that circulating non-classical monocytes are directly recruited to polymer films within skin injuries, where they home to a perivascular niche and generate alternatively activated, wound healing macrophages. Selective labeling of blood monocyte subsets indicates that non-classical monocytes are biased progenitors of alternatively activated macrophages. On-site delivery of the immunomodulatory small molecule FTY720 recruits S1PR3-expressing non-classical monocytes that support vascular remodeling after injury. These results elucidate a previously unknown role for blood-derived non-classical monocytes as contributors to alternatively activated macrophages, highlighting them as key regulators of inflammatory response and regenerative outcome.

Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data.

We present a modular approach for analyzing calcium imaging recordings of large neuronal ensembles. Our goal is to simultaneously identify the locations of the neurons, demix spatially overlapping components, and denoise and deconvolve the spiking activity from the slow dynamics of the calcium indicator. Our approach relies on a constrained nonnegative matrix factorization that expresses the spatiotemporal fluorescence activity as the product of a spatial matrix that encodes the spatial footprint of each neuron in the optical field and a temporal matrix that characterizes the calcium concentration of each neuron over time. This framework is combined with a novel constrained deconvolution approach that extracts estimates of neural activity from fluorescence traces, to create a spatiotemporal processing algorithm that requires minimal parameter tuning. We demonstrate the general applicability of our method by applying it to in vitro and in vivo multi-neuronal imaging data, whole-brain light-sheet imaging data, and dendritic imaging data.

Bayesian Nonparametric Methods for Partially-Observable Reinforcement Learning.

Making intelligent decisions from incomplete information is critical in many applications: for example, robots must choose actions based on imperfect sensors, and speech-based interfaces must infer a user's needs from noisy microphone inputs. What makes these tasks hard is that often we do not have a natural representation with which to model the domain and use for choosing actions; we must learn about the domain's properties while simultaneously performing the task. Learning a representation also involves trade-offs between modeling the data that we have seen previously and being able to make predictions about new data. This article explores learning representations of stochastic systems using Bayesian nonparametric statistics. Bayesian nonparametric methods allow the sophistication of a representation to scale gracefully with the complexity in the data. Our main contribution is a careful empirical evaluation of how representations learned using Bayesian nonparametric methods compare to other standard learning approaches, especially in support of planning and control. We show that the Bayesian aspects of the methods result in achieving state-of-the-art performance in decision making with relatively few samples, while the nonparametric aspects often result in fewer computations. These results hold across a variety of different techniques for choosing actions given a representation.

Decoding arm and hand movements across layers of the macaque frontal cortices.

A major goal for brain machine interfaces is to allow patients to control prosthetic devices with high degrees of independent movements. Such devices like robotic arms and hands require this high dimensionality of control to restore the full range of actions exhibited in natural movement. Current BMI strategies fall well short of this goal allowing the control of only a few degrees of freedom at a time. In this paper we present work towards the decoding of 27 joint angles from the shoulder, arm and hand as subjects perform reach and grasp movements. We also extend previous work in examining and optimizing the recording depth of electrodes to maximize the movement information that can be extracted from recorded neural signals.

Dead leaves and the dirty ground: low-level image statistics in transmissive and occlusive imaging environments.

The opacity of typical objects in the world results in occlusion, an important property of natural scenes that makes inference of the full three-dimensional structure of the world challenging. The relationship between occlusion and low-level image statistics has been hotly debated in the literature, and extensive simulations have been used to determine whether occlusion is responsible for the ubiquitously observed power-law power spectra of natural images. To deepen our understanding of this problem, we have analytically computed the two- and four-point functions of a generalized "dead leaves" model of natural images with parameterized object transparency. Surprisingly, transparency alters these functions only by a multiplicative constant, so long as object diameters follow a power-law distribution. For other object size distributions, transparency more substantially affects the low-level image statistics. We propose that the universality of power-law power spectra for both natural scenes and radiological medical images, formed by the transmission of x-rays through partially transparent tissue, stems from power-law object size distributions, independent of object opacity.