Bayesian Networks in Radiology.

A Bayesian network is a graphical model that uses probability theory to represent relationships among its variables. The model is a directed acyclic graph whose nodes represent variables, such as the presence of a disease or an imaging finding. Connections between nodes express causal influences between variables as probability values. Bayesian networks can learn their structure (nodes and connections) and/or conditional probability values from data. Bayesian networks offer several advantages: (a) they can efficiently perform complex inferences, (b) reason from cause to effect or vice versa, (c) assess counterfactual data, (d) integrate observations with canonical ("textbook") knowledge, and (e) explain their reasoning. Bayesian networks have been employed in a wide variety of applications in radiology, including diagnosis and treatment planning. Unlike deep learning approaches, Bayesian networks have not been applied to computer vision. However, hybrid artificial intelligence systems have combined deep learning models with Bayesian networks, where the deep learning model identifies findings in medical images and the Bayesian network formulates and explains a diagnosis from those findings. One can apply a Bayesian network's probabilistic knowledge to integrate clinical and imaging findings to support diagnosis, treatment planning, and clinical decision-making. This article reviews the fundamental principles of Bayesian networks and summarizes their applications in radiology. Keywords: Bayesian Network, Machine Learning, Abdominal Imaging, Musculoskeletal Imaging, Breast Imaging, Neurologic Imaging, Radiology Education Supplemental material is available for this article. © RSNA, 2023.

Differentiation Between Glioblastoma and Metastatic Disease on Conventional MRI Imaging Using 3D-Convolutional Neural Networks: Model Development and Validation.

RATIONALE AND OBJECTIVES: Imaging-based differentiation between glioblastoma (GB) and brain metastases (BM) remains challenging. Our aim was to evaluate the performance of 3D-convolutional neural networks (CNN) to address this binary classification problem. MATERIALS AND METHODS: T1-CE, T2WI, and FLAIR 3D-segmented masks of 307 patients (157 GB and 150 BM) were generated post resampling, co-registration normalization and semi-automated 3D-segmentation and used for internal model development. Subsequent external validation was performed on 59 cases (27 GB and 32 BM) from another institution. Four different mask-sequence combinations were evaluated using area under the curve (AUC), precision, recall and F1-scores. Diagnostic performance of a neuroradiologist and a general radiologist, both without and with the model output available, was also assessed. RESULTS: 3D-model using the T1-CE tumor mask (TM) showed the highest performance [AUC 0.93 (95% CI 0.858-0.995)] on the external test set, followed closely by the model using T1-CE TM and FLAIR mask of peri-tumoral region (PTR) [AUC of 0.91 (95% CI 0.834-0.986)]. Models using T2WI masks showed robust performance on the internal dataset but lower performance on the external set. Both neuroradiologist and general radiologist showed improved performance with model output provided [AUC increased from 0.89 to 0.968 (p = 0.06) and from 0.78 to 0.965 (p = 0.007) respectively], the latter being statistically significant. CONCLUSION: 3D-CNNs showed robust performance for differentiating GB from BMs, with T1-CE TM, either alone or combined with FLAIR-PTR masks. Availability of model output significantly improved the accuracy of the general radiologist.

AI-based classification of three common malignant tumors in neuro-oncology: A multi-institutional comparison of machine learning and deep learning methods.

PURPOSE: To determine if machine learning (ML) or deep learning (DL) pipelines perform better in AI-based three-class classification of glioblastoma (GBM), intracranial metastatic disease (IMD) and primary CNS lymphoma (PCNSL). METHODOLOGY: Retrospective analysis included 502 cases for training (208 GBM, 67 PCNSL and 227 IMD), with external validation on 86 cases (27:27:32). Multiparametric MRI images (T1W, T2W, FLAIR, DWI and T1-CE) were co-registered, resampled, denoised and intensity normalized, followed by semiautomatic 3D segmentation of the enhancing tumor (ET) and peritumoral region (PTR). Model performance was assessed using several ML pipelines and 3D-convolutional neural networks (3D-CNN) using sequence specific masks, as well as combination of masks. All pipelines were trained and evaluated with 5-fold nested cross-validation on internal data followed by external validation using multi-class AUC. RESULTS: Two ML models achieved similar performance on test set, one using T2-ET and T2-PTR masks (AUC: 0.885, 95% CI: [0.816, 0.935] and another using T1-CE-ET and FLAIR-PTR mask (AUC: 0.878, CI: [0.804, 0.930]). The best performing DL models achieved an AUC of 0.854, (CI [0.774, 0.914]) on external data using T1-CE-ET and T2-PTR masks, followed by model derived from T1-CE-ET, ADC-ET and FLAIR-PTR masks (AUC: 0.851, CI [0.772, 0.909]). CONCLUSION: Both ML and DL derived pipelines achieved similar performance. T1-CE mask was used in three of the top four overall models. Additionally, all four models had some mask derived from PTR, either T2WI or FLAIR.

Transient synaptic potentiation in nucleus accumbens shell during refraining from cocaine seeking.

Repeated exposure to drug-associated cues without reward (extinction) leads to refraining from drug seeking in rodents. We determined if refraining is associated with transient synaptic plasticity (t-SP) in nucleus accumbens shell (NAshell), akin to the t-SP measured in the NAcore during cue-induced reinstatement of drug seeking. Using whole cell patch electrophysiology, we found that medium spiny neurons (MSNs) in NAshell expressed increased ratio of AMPA to NMDA glutamate receptor currents during refraining, which normalized to baseline levels by the end of the 2-hour extinction session. Unlike t-SP observed in NAcore during reinstated drug seeking, neither dendrite spine head enlargement nor activation of matrix metalloproteases (MMP2/9) accompanied the increased AMPA:NMDA in NAshell during refraining. Refraining was also not associated with changes in paired pulse ratio, NMDA receptor current decay time, or AMPA receptor rectification index in NAshell MSNs. Our preliminary data in transgenic mice suggest that t-SP may increase D2-MSN inputs relative to D1-MSN inputs.

Drug Refraining and Seeking Potentiate Synapses on Distinct Populations of Accumbens Medium Spiny Neurons.

Cocaine-associated cues and contexts can precipitate drug seeking in humans and in experimental animals. Glutamatergic synapses in the core subcompartment of the nucleus accumbens (NAcore) undergo transient potentiation in response to presenting drug-associated cues. The NAcore contains two populations of medium spiny neurons (MSNs) that differentially express D1 or D2 dopamine receptors. By recording the ratio of AMPA and NMDA glutamate receptor currents (AMPA/NMDA ratio) from MSNs in NAcore tissue slices, we endeavored to understand which subpopulation of MSNs was undergoing transient potentiation. Transgenic female and male mice differentially expressing fluorescent reporters in D1 or D2 MSNs were withdrawn for 2-3 weeks after being trained to self-administer cocaine. In some mice, discrete cocaine-conditioned cues were isolated from the drug-associated context via extinction training, which causes rodents to refrain from drug seeking in the extinguished context. By measuring AMPA/NMDA ratios in the drug context with or without contextual or discrete cues, and with or without extinction training, we made the following three discoveries: (1) mice refraining from cocaine seeking in the extinguished context showed selective elevation in AMPA/NMDA ratios in D2 MSNs; (2) without extinction training, the drug-associated context selectively increased AMPA/NMDA ratios in D1 MSNs; (3) mice undergoing cue-induced cocaine seeking after extinction training in the drug-associated context showed AMPA/NMDA ratio increases in both D1 and D2 MSNs. These findings reveal that the NAcore codes drug seeking through transient potentiation of D1 MSNs, and that refraining from cocaine seeking in an extinguished context is coded through transient potentiation of D2 MSNs.SIGNIFICANCE STATEMENT Relapse is a primary symptom of addiction that can involve competition between the desire to use drugs and the desire to refrain from using drugs. Drug-associated cues induce relapse, which is correlated with transiently potentiated glutamatergic synapses in the nucleus accumbens core. We determined which of two cell populations in the accumbens core, D1-expressing or D2-expressing neurons, undergo transient synaptic potentiation. After being trained to self-administer cocaine, mice underwent withdrawal, some with and others without extinguishing responding in the drug-associated context. Extinguished mice showed transient potentiation in D2-expressing neurons in the extinguished environment, and all mice engaged in context-induced or cue-induced drug seeking showed transient potentiation of D1-expressing neurons. A simple binary engram in accumbens for seeking drugs and refraining from drugs offers opportunities for cell-specific therapies.

Cocaine Use Reverses Striatal Plasticity Produced During Cocaine Seeking.

BACKGROUND: Relapse is a two-component process consisting of a highly motivated drug-seeking phase that, if successful, is followed by a drug-using phase resulting in temporary satiation. In rodents, cue-induced drug seeking requires transient synaptic potentiation (t-SP) of cortical glutamatergic synapses on nucleus accumbens core medium spiny neurons, but it is unknown how achieving drug use affects this plasticity. We modeled the two phases of relapse after extinction from cocaine self-administration to assess how cocaine use affects t-SP associated with cue-induced drug seeking. METHODS: Rats were trained to self-administer cocaine (n = 96) or were used as yoked-saline control animals (n = 21). After extinction, reinstatement was initiated by 10 minutes of cue-induced drug seeking, followed by 45 minutes with contingent cocaine access, after which cocaine was discontinued and unreinforced lever pressing ensued. Three measures of t-SP were assayed during reinstatement: dendritic spine morphology, alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) to N-methyl-D-aspartate (NMDA) ratios, and matrix metalloproteinase activity. RESULTS: We found that cocaine use for 10 minutes collapsed all three measures of cue-potentiated t-SP back to baseline. Moreover, when cocaine use was discontinued 45 minutes later, dendritic spine morphology and AMPA to NMDA ratios were restored as animals became motivated to engage unrewarded lever pressing. Nonreinforced drug seeking was positively correlated with changes in spine morphology, and cocaine access reversed this relationship. CONCLUSIONS: Using a novel modification of the reinstatement paradigm, we show that achieving cocaine use reversed the synaptic plasticity underpinning the motivation to seek the drug.

Glutamate Transporter GLT-1 as a Therapeutic Target for Substance Use Disorders.

The development of new treatments for substance use disorders requires identification of targetable molecular mechanisms. Pathology in glutamatergic neurotransmission system in brain reward circuitry has been implicated in relapse to multiple classes of drugs. Glutamate transporter 1 (GLT-1) crucially regulates glutamatergic signaling by removing excess glutamate from the extrasynaptic space. The purpose of this review is to highlight the effects of addictive drug use on GLT-1 and glutamate uptake, and using GLT-1 as a target in addiction pharmacotherapy. Cocaine, opioids, ethanol, nicotine, amphetamines, and cannabinoids each affect GLT-1 expression and glutamate uptake, and restoring GLT-1 expression with N-acetylcysteine or ceftriaxone shows promise in correcting pre-clinical and clinical manifestations of drug addiction.

Mindfulness training alters emotional memory recall compared to active controls: support for an emotional information processing model of mindfulness.

OBJECTIVES: While mindfulness-based interventions have received widespread application in both clinical and non-clinical populations, the mechanism by which mindfulness meditation improves well-being remains elusive. One possibility is that mindfulness training alters the processing of emotional information, similar to prevailing cognitive models of depression and anxiety. The aim of this study was to investigate the effects of mindfulness training on emotional information processing (i.e., memory) biases in relation to both clinical symptomatology and well-being in comparison to active control conditions. METHODS: Fifty-eight university students (28 female, age = 20.1 ± 2.7 years) participated in either a 12-week course containing a "meditation laboratory" or an active control course with similar content or experiential practice laboratory format (music). Participants completed an emotional word recall task and self-report questionnaires of well-being and clinical symptoms before and after the 12-week course. RESULTS: Meditators showed greater increases in positive word recall compared to controls [F(1, 56) = 6.6, p = 0.02]. The meditation group increased significantly more on measures of well-being [F(1, 56) = 6.6, p = 0.01], with a marginal decrease in depression and anxiety [F(1, 56) = 3.0, p = 0.09] compared to controls. Increased positive word recall was associated with increased psychological well-being (r = 0.31, p = 0.02) and decreased clinical symptoms (r = -0.29, p = 0.03). CONCLUSION: Mindfulness training was associated with greater improvements in processing efficiency for positively valenced stimuli than active control conditions. This change in emotional information processing was associated with improvements in psychological well-being and less depression and anxiety. These data suggest that mindfulness training may improve well-being via changes in emotional information processing. Future research with a fully randomized design will be needed to clarify the possible influence of self-selection.