A BERT-GNN Approach for Metastatic Breast Cancer Prediction Using Histopathology Reports.

Metastatic breast cancer (MBC) continues to be a leading cause of cancer-related deaths among women. This work introduces an innovative non-invasive breast cancer classification model designed to improve the identification of cancer metastases. While this study marks the initial exploration into predicting MBC, additional investigations are essential to validate the occurrence of MBC. Our approach combines the strengths of large language models (LLMs), specifically the bidirectional encoder representations from transformers (BERT) model, with the powerful capabilities of graph neural networks (GNNs) to predict MBC patients based on their histopathology reports. This paper introduces a BERT-GNN approach for metastatic breast cancer prediction (BG-MBC) that integrates graph information derived from the BERT model. In this model, nodes are constructed from patient medical records, while BERT embeddings are employed to vectorise representations of the words in histopathology reports, thereby capturing semantic information crucial for classification by employing three distinct approaches (namely univariate selection, extra trees classifier for feature importance, and Shapley values to identify the features that have the most significant impact). Identifying the most crucial 30 features out of 676 generated as embeddings during model training, our model further enhances its predictive capabilities. The BG-MBC model achieves outstanding accuracy, with a detection rate of 0.98 and an area under curve (AUC) of 0.98, in identifying MBC patients. This remarkable performance is credited to the model's utilisation of attention scores generated by the LLM from histopathology reports, effectively capturing pertinent features for classification.

Bayesian Activity Estimation and Uncertainty Quantification of Spent Nuclear Fuel Using Passive Gamma Emission Tomography.

In this paper, we address the problem of activity estimation in passive gamma emission tomography (PGET) of spent nuclear fuel. Two different noise models are considered and compared, namely, the isotropic Gaussian and the Poisson noise models. The problem is formulated within a Bayesian framework as a linear inverse problem and prior distributions are assigned to the unknown model parameters. In particular, a Bernoulli-truncated Gaussian prior model is considered to promote sparse pin configurations. A Markov chain Monte Carlo (MCMC) method, based on a split and augmented Gibbs sampler, is then used to sample the posterior distribution of the unknown parameters. The proposed algorithm is first validated by simulations conducted using synthetic data, generated using the nominal models. We then consider more realistic data simulated using a bespoke simulator, whose forward model is non-linear and not available analytically. In that case, the linear models used are mis-specified and we analyse their robustness for activity estimation. The results demonstrate superior performance of the proposed approach in estimating the pin activities in different assembly patterns, in addition to being able to quantify their uncertainty measures, in comparison with existing methods.

Bayesian bacterial detection using irregularly sampled optical endomicroscopy images.

Pneumonia is a major cause of morbidity and mortality of patients in intensive care. Rapid determination of the presence and gram status of the pathogenic bacteria in the distal lung may enable a more tailored treatment regime. Optical Endomicroscopy (OEM) is an emerging medical imaging platform with preclinical and clinical utility. Pulmonary OEM via multi-core fibre bundles has the potential to provide in vivo, in situ, fluorescent molecular signatures of the causes of infection and inflammation. This paper presents a Bayesian approach for bacterial detection in OEM images. The model considered assumes that the observed pixel fluorescence is a linear combination of the actual intensity value associated with tissues or background, corrupted by additive Gaussian noise and potentially by an additional sparse outlier term modelling anomalies (bacteria). The bacteria detection problem is formulated in a Bayesian framework and prior distributions are assigned to the unknown model parameters. A Markov chain Monte Carlo algorithm based on a partially collapsed Gibbs sampler is used to sample the posterior distribution of the unknown parameters. The proposed algorithm is first validated by simulations conducted using synthetic datasets for which good performance is obtained. Analysis is then conducted using two ex vivo lung datasets in which fluorescently labelled bacteria are present in the distal lung. A good correlation between bacteria counts identified by a trained clinician and those of the proposed method, which detects most of the manually annotated regions, is observed.

Characterization and modelling of inter-core coupling in coherent fiber bundles.

Recent developments in optical endomicroscopy (OEM) and associated fluorescent SmartProbes present a need for sensitive imaging with high detection performance. Inter-core coupling within coherent fiber bundles is a well recognized limitation, affecting the technology's imaging capabilities. Fiber cross coupling has been studied both experimentally and within a theoretical framework (coupled mode theory), providing (i) insights on the factors affecting cross talk, and (ii) recommendations for optimal fiber bundle design. However, due to physical limitations, such as the tradeoff between cross coupling and core density, cross coupling can be suppressed yet not eliminated through optimal fiber design. This study introduces a novel approach for measuring, analyzing and quantifying cross coupling within coherent fiber bundles, in a format that can be integrated into a linear model, which in turn can enable computational compensation of the associated blurring introduced to OEM images.

Influence of the analysis technique on estimating hepatic iron content using MRI.

PURPOSE: To investigate the effect of the analysis technique on estimating hepatic iron content using MRI. MATERIALS AND METHODS: We evaluated the influences of single-exponential (EXP), bi-exponential (BEXP), and exponential-plus-constant (CEXP) models; and pixel-wise (MAP), average (AVG), and median (MED) signal calculation methods on T2* measurement using numerical simulations, calibrated phantoms, and nine patients scanned on 3 Tesla MRI, based on regression, correlation, and t-test statistical analysis. RESULTS: The T2* measurement error varied from 9 to 51% in the numerical simulations (T2*: 5-20 ms), depending on signal-to-noise ratio (SNR; range: 8-233) with significant (P < 0.05) difference between actual and predicted values. The MAP method performed well (error < 10%) at high SNR (>100), but resulted in severe estimation errors at low SNR (<50). The EXP model resulted in significant measurement differences (P < 0.05) compared with all other methods, irrespective of SNR. In vivo T2* values ranged from 3.1 to 53.6 ms, depending on the amount of iron overload and implemented analysis method. The BEXP (range: 3.7-50 ms) and CEXP (range: 3.8-53.6 ms) models, and the AVG (range: 3.2-38.8 ms) and MED (range: 3.1-38.5 ms) methods provided more accurate measurements than the EXP model (range: 3.1-18.3 ms) and MAP (range: 3.8-53.6 ms) method, respectively (P < 0.05). The BEXP and CEXP models provided very similar measurements (P > 0.87). Similarly, the AVG and MED methods provided very similar results (P > 0.97), with slightly better performance of the AVG method. CONCLUSION: Different analysis techniques show different performances based on the fitting model and signal calculation method. Based on this study, the CEXP model and AVG method are recommended due to simpler implementation and less influence by the selected analysis region. J. Magn. Reson. Imaging 2016;44:1448-1455.

MRI T2* imaging for assessment of liver iron overload: study of different data analysis approaches.

Background Recently, magnetic resonance imaging (MRI) has been established as an effective technique for evaluating iron overload by measuring T2* in the liver. Purpose To investigate the effects of various factors associated with T2* calculation on the resulting measurement and to determine the analysis criterion that provides the most accurate T2* measurements. Material and Methods Both phantom and in vivo MRI experiments were conducted to study the effects of the selected region of interest (ROI) location and size, signal-averaging method, exponential-fitting model, echo truncation, iron-overload severity, and inter-/intra-observer variabilities on T2* measurements. The results were compared to reference values from the scanner processing software. Results The pixel-by-pixel calculation method provided results in better agreement with the reference values from the MRI scanner than the average or median methods. The choice of the exponential fitting model affected the results, depending on signal-to-noise ratio, number of echoes, minimum and maximum echo times, and tissue composition inside the selected ROI. The single-exponential model resulted in smaller error than the bi-exponential or exponential-plus-constant models, where the latter two models showed similar results. The relative performance of the different models and methods was not affected by the degree of iron-overload. Conclusion Various factors associated with the adopted T2* calculation method affect the resulting measurement. In this study, the pixel-by-pixel calculation method and single-exponential model provided the most accurate results based on the conducted phantom and in vivo MRI experiments.

The influence of the analysis technique on estimating liver iron overload using magnetic resonance imaging T2* quantification.

Iron toxicity is the major cause of tissue damage in patients with iron overload. Iron deposits mainly in the liver, where its concentration closely correlates with whole body iron overload. Different techniques have been proposed for estimating iron content, with liver biopsy being the gold standard despite its invasiveness and influence by sampling error. Recently, magnetic resonance imaging (MRI) has been established as an effective technique for evaluating iron overload by measuring T2(*) in the liver. However, various factors associated with the adopted analysis technique, mainly the exponential fitting model and signal averaging method, affect the resulting measurements. In this study, we evaluate the influences of these factors on T2(*) measurement in numerical phantom, calibrated phantoms, and nine patients with different degrees of iron overload. The results show different performances among the fitting models and signal averaging methods, which are affected by SNR, image quality and signal homogeneity inside the selected ROI for analysis.

Deep neck infections of congenital causes.

OBJECTIVES: To review cases of deep neck infections with underlying congenital etiology with special emphasis on their clinical presentations and the computed tomographic findings and to discuss the various therapeutic modalities employed for such lesions. STUDY DESIGN: Case series with chart review. SETTINGS: Alexandria University Hospital, Egypt. SUBJECTS AND METHODS: The authors retrospectively reviewed the clinical, imaging, and operative records of deep neck infection cases presented to their department in the past 10 years. Deep neck infection cases due to congenital causes were included in the study. RESULTS Of the 249 cases of deep neck infections admitted to the authors' department in the past 10 years, 39 patients were diagnosed with deep neck infections due to congenital causes. Patients were classified into 2 groups. In group 1 (29 patients), computed tomography revealed the presence of infected cystic swelling in the neck that was classified as second branchial cyst (16 patients), third and fourth branchial cysts (8 patients), and thyroglossal cyst (5 patients). Group 2 (10 patients) presented with recurrent attacks of deep neck infection with a history of incision and drainage several times. Radiological and operative findings revealed the presence of congenital pyriform fossa sinus. CONCLUSION: Computed tomography is helpful in diagnosing infected congenital cysts and its types. Infected congenital cysts could be excised completely under an umbrella of antibiotics. Recurrence of deep neck infections should alert the physician to the possibility of underlying congenital lesions. Thorough clinical and radiological assessment is mandatory to rule out the possibility of a congenital pyriform fossa sinus.

Temporalis myofascial flap for primary cranial base reconstruction after tumor resection.

OBJECTIVE: To evaluate the use of the temporalis myofascial flap in primary cranial base reconstruction following surgical tumor ablation and to explain technical issues, potential complications, and donor site consequences along with their management. DESIGN: Retrospective case series. SETTING: Tertiary referral center. PARTICIPANTS: Forty-one consecutive patients receiving primary temporalis myofascial flap reconstructions following cranial base tumor resections in a 4-year period. MAIN OUTCOME MEASURES: Flap survival, postoperative complications, and donor site morbidity. RESULTS: Patients included 37 males and 4 females ranging in age from 10 to 65 years. Two patients received preoperative and 18 postoperative radiation therapy. Patient follow-up ranged from 4 to 39 months. The whole temporalis muscle was used in 26 patients (63.4%) and only part of a coronally split muscle was used in 15 patients (36.6%). Nine patients had primary donor site reconstruction using a Medpor((R)) (Porex Surgical, Inc., Newnan, GA) temporal fossa implant; these had excellent aesthetic results. There were no cases of complete flap loss. Partial flap dehiscence was seen in six patients (14.6%); only two required surgical débridement. None of the patients developed cerebrospinal leaks or meningitis. One patient was left with complete paralysis of the temporal branch of the facial nerve. Three patients (all had received postoperative irradiation) developed permanent trismus. CONCLUSIONS: The temporalis myofascial flap was found to be an excellent reconstructive alternative for a wide variety of skull base defects following tumor ablation. It is a very reliable, versatile flap that is usually available in the operative field with relatively low donor site aesthetic and functional morbidity.