Data Quality Estimation Via Model Performance: Machine Learning as a Validation Tool.

In our recent study, the attempt to classify neurosurgical operative reports into routinely used expert-derived classes exhibited an F-score not exceeding 0.74. This study aimed to test how improving the classifier (target variable) affected the short text classification with deep learning on real-world data. We redesigned the target variable based on three strict principles when applicable: pathology, localization, and manipulation type. The deep learning significantly improved with the best result of operative report classification into 13 classes (accuracy = 0.995, F1 = 0.990). Reasonable text classification with machine learning should be a two-way process: the model performance must be ensured by the unambiguous textual representation reflected in corresponding target variables. At the same time, the validity of human-generated codification can be inspected via machine learning.

Compact ring resonators of silicon nanorods for strong optomechanical interaction.

Optomechanical interaction in microstructures plays a more and more important role in the fields of quantum technology, information processing, and sensing, among others. It is still a challenge to obtain a strong optomechanical interaction in a compact device. Here, we propose and demonstrate that compact ring resonators consisting of silicon nanorods can realize strong optomechanical interaction even surpassing that of most optical microcavities. The proposed ring resonators can well confine infrared optical waves by the quasi-bound states in the continuum. Meanwhile, each nanorod in the resonator acts as a mechanical resonator of GHz resonating frequency, thus realizing an optomechanical coupling rate of up to 1.8 MHz. We have found that the interaction area can be extended by increasing the number of nanorods while maintaining the optomechanical interaction strength. Finally, we have studied the influence of supporting structures for suspended nanorods on the optomechanical interaction properties. The proposed ring resonators of silicon nanorods offer a promising platform for the study of optomechanical interaction.

5-ALA-guided tumor resection during awake speech mapping in gliomas located in eloquent speech areas: Single-center experience.

Background: Achieving maximal functionally safe resection of gliomas located within the eloquent speech areas is challenging, and there is a lack of literature on the combined use of 5-aminolevulinic acid (5-ALA) guidance and awake craniotomy. Objective: The aim of this study was to describe our experience with the simultaneous use of 5-ALA fluorescence and awake speech mapping in patients with left frontal gliomas located within the vicinity of eloquent speech areas. Materials and methods: A prospectively collected database of patients was reviewed. 5-ALA was administered at a dose of 20 mg/kg 2 h prior to operation, and an operating microscope in BLUE400 mode was used to visualize fluorescence. All patients underwent surgery using the "asleep-awake-asleep" protocol with monopolar and bipolar electrical stimulation to identify the proximity of eloquent cortex and white matter tracts and to guide safe limits of resection along with fluorescence guidance. Speech function was assessed by a trained neuropsychologist before, during, and after surgery. Results: In 28 patients operated with cortical mapping and 5-ALA guidance (12 Grade 4, 6 Grade 3, and 10 Grade 2 gliomas), Broca's area was identified in 23 cases and Wernicke's area was identified in 5 cases. Fluorescence was present in 14 cases. Six tumors had residual fluorescence due to the positive speech mapping in the tumor bed. Transient aphasia developed in 14 patients, and permanent aphasia developed in 4 patients. In 6 patients operated with cortical and subcortical speech mapping and 5-ALA guidance (4 Grade 4, 1 Grade 3, and 1 Grade 2 gliomas), cortical speech areas were mapped in 5 patients and subcortical tracts were encountered in all cases. In all cases, resection was stopped despite the presence of residual fluorescence due to speech mapping findings. Transient aphasia developed in 6 patients and permanent aphasia developed in 4 patients. In patients with Grade 2-3 gliomas, targeted biopsy of focal fluorescence areas led to upgrading the grade and thus more accurate diagnosis. Conclusion: 5-ALA guidance during awake speech mapping is useful in augmenting the extent of resection for infiltrative high-grade gliomas and identifying foci of anaplasia in non-enhancing gliomas, while maintaining safe limits of functional resection based on speech mapping. Positive 5-ALA fluorescence in diffuse Grade 2 gliomas may be predictive of a more aggressive disease course.

Reparative properties of human glioblastoma cells after single exposure to a wide range of X-ray doses.

Radiation therapy induces double-stranded DNA breaks in tumor cells, which leads to their death. A fraction of glioblastoma cells repair such breaks and reinitiate tumor growth. It was necessary to identify the relationship between high radiation doses and the proliferative activity of glioblastoma cells, and to evaluate the contribution of DNA repair pathways, homologous recombination (HR), and nonhomologous end joining (NHEJ) to tumor-cell recovery. We demonstrated that the GO1 culture derived from glioblastoma cells from Patient G, who had previously been irradiated, proved to be less sensitive to radiation than the Sus\fP2 glioblastoma culture was from Patient S, who had not been exposed to radiation before. GO1 cell proliferation decreased with radiation dose, and MTT decreased to 35% after a single exposure to 125 Gγ. The proliferative potential of glioblastoma culture Sus\fP2 decreased to 35% after exposure to 5 Gγ. At low radiation doses, cell proliferation and the expression of RAD51 were decreased; at high doses, cell proliferation was correlated with Ku70 protein expression. Therefore, HR and NHEJ are involved in DNA break repair after exposure to different radiation doses. Low doses induce HR, while higher doses induce the faster but less accurate NHEJ pathway of double-stranded DNA break repair.

Multinomial Classification of Neurosurgical Operations Using Gradient Boosting and Deep Learning Algorithms.

This study aimed at testing the feasibility of neurosurgical procedures classification into 100+ classes using natural language processing and machine learning. A catboost algorithm and bidirectional recurrent neural network with a gated recurrent unit showed almost the same accuracy of ∼81%, with suggestions of correct class in top 2-3 scored classes up to 98.9%. The classification of neurosurgical procedures via machine learning appears to be a technically solvable task which can be additionally improved considering data enhancement and classes verification.

Predicting the Length of Stay in Neurosurgery with RuGPT-3 Language Model.

In this study, we update the evaluation of the Russian GPT3 model presented in our previous paper in predicting the length of stay (LOS) in neurosurgery. We aimed to assess the performance the Russian GPT-3 (ruGPT-3) language model in LOS prediction using narrative medical records in neurosurgery compared to doctors' and patients' expectations. Doctors appeared to have the most realistic LOS expectations (MAE = 2.54), while the model's predictions (MAE = 3.53) were closest to the patients' (MAE = 3.47) but inferior to them (p = 0.011). A detailed analysis showed a solid quality of ruGPT-3 performance based on narrative clinical texts. Considering our previous findings obtained with recurrent neural networks and FastText vector representation, we estimate the new result as important but probably improveable.

The Classification of Scientific Abstracts Using Text Statistical Features.

Automated abstracts classification could significantly facilitate scientific literature screening. The classification of short texts could be based on their statistical properties. This research aimed to evaluate the quality of short medical abstracts classification primarily based on text statistical features. Twelve experiments with machine learning models over the sets of text features were performed on a dataset of 671 article abstracts. Each experiment was repeated 300 times to estimate the classification quality, ending up with 3600 tests total. We achieved the best result (F1 = 0.775) using a random forest machine learning model with keywords and three-dimensional Word2Vec embeddings. The classification of scientific abstracts might be implemented using straightforward and computationally inexpensive methods presented in this paper. The approach we described is expected to facilitate literature selection by researchers.

Noninvasive Glioma Grading with Deep Learning: A Pilot Study.

Gliomas are the most common neuroepithelial brain tumors, different by various biological tissue types and prognosis. They could be graded with four levels according to the 2007 WHO classification. The emergence of non-invasive histological and molecular diagnostics for nervous system neoplasms can revolutionize the efficacy and safety of medical care and radically reduce healthcare costs. Our pilot study aimed to evaluate the diagnostic accuracy of deep learning (DL) in subtyping gliomas by WHO grades (I-IV) based on preoperative magnetic resonance imaging (MRI) from Burdenko Neurosurgery Center's database. A total of 707 MRI studies was included. A "3D classification" approach predicting tumor type for the entire patient's MRI data showed the best result (accuracy = 83%, ROC AUC = 0.95), consistent with that of other authors who used different methodologies. Our preliminary results proved the separability of MR T1 axial images with contrast enhancement by WHO grade using DL.

First-in-Man Noninvasive Initial Diagnostic Approach of Primary CNS Lymphoma Versus Glioblastoma Using PET With 18 F-Fludarabine and l -[methyl- 11 C]Methionine.

OBJECTIVES: This study sought to assess 18 F-fludarabine ( 18 F-FLUDA) PET/CT's ability in differentiating primary central nervous system lymphomas (PCNSLs) from glioblastoma multiformes (GBMs). PATIENTS AND METHODS: Patients harboring either PCNSL (n = 8) before any treatment, PCNSL treated using corticosteroids (PCNSLh; n = 10), or GBM (n = 13) were investigated with conventional MRI and PET/CT, using 11 C-MET and 18 F-FLUDA. The main parameters measured with each tracer were SUV T and T/N ratios for the first 30 minutes of 11 C-MET acquisition, as well as at 3 different times after 18 F-FLUDA injection. The early 18 F-FLUDA uptake within the first minute of injection was equally considered, whereas this parameter was combined with the later uptakes to obtain R FLUDA 2 and R FLUDA 3 ratios. RESULTS: No significant differences in 11 C-MET uptakes were observed among PCNSL, PCNSLh, and GBM. With 18 F-FLUDA, a clear difference in dynamic GBM uptake was observed, which decreased over time after an early maximum, as compared with that of PCNSL, which steadily increased over time, PCNSLh exhibiting intermediate values. The most discriminative parameters consisting of R FLUDA 2 and R FLUDA 3 integrated the early tracer uptake (first 60 seconds), thereby provided 100% specificity and sensitivity. CONCLUSIONS: 18 F-FLUDA was shown to likely be a promising radiopharmaceutical for differentiating PCNSL from other malignancies, although a pretreatment with corticosteroids might compromise this differential diagnostic ability. The diagnostic role of 18 F-FLUDA should be further investigating, along with its potential of defining therapeutic strategies in patients with PCNSL, while assessing the treatments' effectiveness.

A Comparison of Word Embeddings to Study Complications in Neurosurgery.

Our study aimed to compare the capability of different word embeddings to capture the semantic similarity of clinical concepts related to complications in neurosurgery at the level of medical experts. Eighty-four sets of word embeddings (based on Word2vec, GloVe, FastText, PMI, and BERT algorithms) were benchmarked in a clustering task. FastText model showed the best close to the medical expertise capability to group medical terms by their meaning (adjusted Rand index = 0.682). Word embedding models can accurately reflect clinical concepts' semantic and linguistic similarities, promising their robust usage in medical domain-specific NLP tasks.

Semiautomatic Identification of Pulmonary Embolism in Electronic Health Records Through Sentence Labeling.

In this study, we tested the quality of the information extraction algorithm proposed by our group to detect pulmonary embolism (PE) in medical cases through sentence labeling. Having shown a comparable result (F1 = 0.921) to the best machine learning method (random forest, F1 = 0.937), our approach proved not to miss the information of interest. Scoping the number of texts under review down to distinct sentences and introducing labeling rules contributes to the efficiency and quality of information extraction by experts and makes the challenging tasks of labeling large textual datasets solvable.

Length of Stay Prediction in Neurosurgery with Russian GPT-3 Language Model Compared to Human Expectations.

Patients, relatives, doctors, and healthcare providers anticipate the evidence-based length of stay (LOS) prediction in neurosurgery. This study aimed to assess the quality of LOS prediction with the GPT3 language model upon the narrative medical records in neurosurgery comparing to doctors' and patients' expectations. We found no significant difference (p = 0.109) between doctors', patients', and model's predictions with neurosurgeons tending to be more accurate in prognosis. The modern neural network language models demonstrate feasibility in LOS prediction.

Best Research Practice Implementation: The Experience of the N.N. Burdenko National Medical Research Center of Neurosurgery.

Implementing the best research principles initiates an important shift in clinical research culture, improving efficiency and the level of evidence obtained. In this article, we share our own view on the best research practice and our experience introducing it into the scientific activities of the N.N. Burdenko National Medical Research Center of Neurosurgery (Moscow, Russian Federation). While being adherent to the principles described in the article, the percentage of publications in the international scientific journals in our Center has increased from 7% to 27%, with an overall gain in the number of articles by 2 times since 2014. We believe it is important that medical informatics professionals equally to medical experts involved in clinical research are familiar with the best research principles.

Strong optomechanical coupling in chain-like waveguides of silicon nanoparticles with quasi-bound states in the continuum.

We propose and demonstrate that strong optomechanical coupling can be achieved in a chain-like waveguide consisting of silicon nanorods. By employing quasi-bound states in the continuum and mechanical resonances at a frequency around 10 GHz, the optomechanical coupling rate can be above 2 MHz and surpass most microcavities. We have also studied cases with different optical wave numbers and size parameters of silicon, and a robust coupling rate has been verified, benefiting the experimental measurements and practical applications. The proposed silicon chain-like waveguide of strong optomechanical coupling may pave new ways for research on photon-phonon interaction with microstructures.

Prediction of Intraoperative Fluorescence of Brain Gliomas: Correlation between Tumor Blood Flow and the Fluorescence.

INTRODUCTION: The prediction of the fluorescent effect of 5-aminolevulinic acid (5-ALA) in patients with diffuse gliomas can improve the selection of patients. The degree of enhancement of gliomas has been reported to predict 5-ALA fluorescence, while, at the same time, rarer cases of fluorescence have been described in non-enhancing gliomas. Perfusion studies, in particular arterial spin labeling perfusion, have demonstrated high efficiency in determining the degree of malignancy of brain gliomas and may be better for predicting fluorescence than contrast enhancement. The aim of the study was to investigate the relationship between tumor blood flow, measured by ASL, and intraoperative fluorescent glow of gliomas of different grades. MATERIALS AND METHODS: Tumoral blood flow was assessed in 75 patients by pCASL (pseudo-continuous arterial spin labeling) within 1 week prior to surgery. In all cases of tumor removal, 5-ALA had been administered preoperatively. Maximum values of tumoral blood flow (TBF max) were measured, and normalized tumor blood flow (nTBF) was calculated. RESULTS: A total of 76% of patients had significant contrast enhancement, while 24% were non-enhancing. The histopathology revealed 17 WHO grade II gliomas, 12 WHO grade III gliomas and 46 glioblastomas. Overall, there was a relationship between the degree of intraoperative tumor fluorescence and ASL-TBF (Rs = 0.28, p = 0.02 or the TBF; Rs = 0.34, p = 0.003 for nTBF). Non-enhancing gliomas were fluorescent in 9/18 patients, with nTBF in fluorescent gliomas being 54.58 ± 32.34 mL/100 mg/s and in non-fluorescent gliomas being 52.99 ± 53.61 mL/100 g/s (p > 0.05). Enhancing gliomas were fluorescent in 53/57 patients, with nTBF being 170.17 ± 107.65 mL/100 g/s in fluorescent and 165.52 ± 141.71 in non-fluorescent gliomas (p > 0.05). CONCLUSION: Tumoral blood flow levels measured by non-contrast ASL perfusion method predict the fluorescence by 5-ALA; however, the additional value beyond contrast enhancement is not clear. ASL is, however, useful in cases with contraindication to contrast.

The Classification of Short Scientific Texts Using Pretrained BERT Model.

Automated text classification is a natural language processing (NLP) technology that could significantly facilitate scientific literature selection. A specific topical dataset of 630 article abstracts was obtained from the PubMed database. We proposed 27 parametrized options of PubMedBERT model and 4 ensemble models to solve a binary classification task on that dataset. Three hundred tests with resamples were performed in each classification approach. The best PubMedBERT model demonstrated F1-score = 0.857 while the best ensemble model reached F1-score = 0.853. We concluded that the short scientific texts classification quality might be improved using the latest state-of-art approaches.

Inter-Rater Reliability of Unstructured Text Labeling: Artificially vs. Naturally Intelligent Approaches.

Unstructured medical text labeling technologies are expected to be highly demanded since the interest in artificial intelligence and natural language processing arises in the medical domain. Our study aimed to assess the agreement between experts who judged on the fact of pulmonary embolism (PE) in neurosurgical cases retrospectively based on electronic health records and assess the utility of the machine learning approach to automate this process. We observed a moderate agreement between 3 independent raters on PE detection (Light's kappa = 0.568, p = 0). Labeling sentences with the method we proposed earlier might improve the machine learning results (accuracy = 0.97, ROC AUC = 0.98) even in those cases that could not be agreed between 3 independent raters. Medical text labeling techniques might be more efficient when strict rules and semi-automated approaches are implemented. Machine learning might be a good option for unstructured text labeling when the reliability of textual data is properly addressed. This project was supported by the RFBR grant 18-29-22085.

Effects of Hyperthermia on Intracranial Pressure and Cerebral Autoregulation in Patients with an Acute Brain Injury.

Hyperthermia is a common detrimental condition in patients with an acute brain injury (ABI), which can worsen their prognosis and outcome. The aim of this study was to evaluate the effects of hyperthermia on intracranial pressure (ICP) and cerebral autoregulation (CA).Eight patients with ABI were studied. CA was assessed on the basis of the pressure reactivity index (PRx) coefficient. The ICP, cerebral perfusion pressure (CPP), and PRx were compared before and during development of hyperthermia. Hyperthermia was defined as an increase in cerebral temperature above 38.3 °C.Thirty-three episodes of hyperthermia were analyzed: 25 of these occurred on a background of initially normal ICP whereas 8 occurred on a background of initially elevated ICP, and 17 of the 33 episodes occurred on a background of initially intact autoregulation whereas 16 occurred on a background of initially impaired autoregulation.During hyperthermia, elevated ICP was found in 52% of instances where it was initially normal, and further progression of intracranial hypertension occurred in 100% of instances where ICP was initially elevated. The median ICP during hyperthermia was 24 [range quartiles 22-28] mmHg in instances where it was initially normal and 31 [quartiles 27-32] mmHg in instances where it was initially elevated (p < 0.01). The correlation coefficient between the brain temperature and ICP was 0.11 (p < 0.01). During hyperthermia, the number of episodes of ICP >20 mmHg increased by 41% in instances with intact autoregulation but ICP was above 20 mmHg and by 38% (p > 0.05) in instances with impaired autoregulation and ICP was 20 mmHg. The cerebral hyperthermia-associated increase in ICP was not associated with impaired autoregulation.

Comparison of Dimensionality Reduction Methods in Mass Spectra of Astrocytoma and Glioblastoma Tissues.

Recently developed methods of ambient ionization allow the collection of mass spectrometric datasets for biological and medical applications at an unprecedented pace. One of the areas that could employ such analysis is neurosurgery. The fast in situ identification of dissected tissues could assist the neurosurgery procedure. In this paper tumor tissues of astrocytoma and glioblastoma are compared. The vast majority of the data representation methods are hard to use, as the number of features is high and the amount of samples is limited. Furthermore, the ratio of features and samples number restricts the use of many machine learning methods. The number of features could be reduced through feature selection algorithms or dimensionality reduction methods. Different algorithms of dimensionality reduction are considered along with the traditional noise thresholding for the mass spectra. From our analysis, the Isomap algorithm appears to be the most effective dimensionality reduction algorithm for negative mode, whereas the positive mode could be processed with a simple noise reduction by a threshold. Also, negative and positive mode correspond to different sample properties: negative mode is responsible for the inner variability and the details of the sample, whereas positive mode describes measurement in general.

Rapid estimation of tumor cell percentage in brain tissue biopsy samples using inline cartridge extraction mass spectrometry.

Tumor cell percentage (TCP) is an essential characteristic of biopsy samples that directly affects the sensitivity of molecular testing in clinical practice. Apart from clarifying diagnoses, rapid evaluation of TCP combined with various neuronavigation systems can be used to support decision making in neurosurgery. It is known that ambient mass spectrometry makes it possible to rapidly distinguish healthy from malignant tissues. In connection with this, here we demonstrate the possibility of using non-imaging ambient mass spectrometry to evaluate TCP in glial tumor tissues with a high degree of confidence. Molecular profiles of histologically annotated human glioblastoma tissue samples were obtained using the inline cartridge extraction ambient mass spectrometry approach. XGBoost regressors were trained to evaluate tumor cell percentage. Using cross-validation, it was estimated that the TCP was determined by the regressors with a precision of approximately 90% using only low-resolution data. This result demonstrates that ambient mass spectrometry provides an accurate method todetermine TCP in dissected tissues even without implementing mass spectrometry imaging. The application of such techniques offers the possibility to automate routine tissue screening and TCP evaluation to boost the throughput of pathology laboratories. Rapid estimation of tumor cell percentage during neurosurgery.