BioBERTurk: Exploring Turkish Biomedical Language Model Development Strategies in Low-Resource Setting.

Pretrained language models augmented with in-domain corpora show impressive results in biomedicine and clinical Natural Language Processing (NLP) tasks in English. However, there has been minimal work in low-resource languages. Although some pioneering works have shown promising results, many scenarios still need to be explored to engineer effective pretrained language models in biomedicine for low-resource settings. This study introduces the BioBERTurk family and four pretrained models in Turkish for biomedicine. To evaluate the models, we also introduced a labeled dataset to classify radiology reports of head CT examinations. Two parts of the reports, impressions and findings, are evaluated separately to observe the performance of models on longer and less informative text. We compared the models with the Turkish BERT (BERTurk) pretrained with general domain text, multilingual BERT (mBERT), and LSTM+attention-based baseline models. The first model initialized from BERTurk and then further pretrained with biomedical corpus performs statistically better than BERTurk, multilingual BERT, and baseline for both datasets. The second model continues to pretrain the BERTurk model by using only radiology Ph.D. theses to test the effect of task-related text. This model slightly outperformed all models on the impression dataset and showed that using only radiology-related data for continual pre-training could be effective. The third model continues to pretrain by adding radiology theses to the biomedical corpus but does not show a statistically meaningful difference for both datasets. The final model combines radiology and biomedicine corpora with the corpus of BERTurk and pretrains a BERT model from scratch. This model is the worst-performing model of the BioBERT family, even worse than BERTurk and multilingual BERT.

Reflections of the sensory findings in the central nervous system in patients with neuropathic pain.

This study aimed to evaluate whether there was a difference in functional magnetic resonance imaging (fMRI) findings in patients who were found having hyperalgesia or hypoesthesia according to Quantitative Sensory Tests (QST). Forty participants were included in the study: 20 with neuropathic pain (NP) due to cervical disc pathology (NP group) and 20 healthy volunteers. After obtaining the socio-demographic and clinical data of the participants, the painDETECT questionnaire was administered, followed by QST analysis to show the presence of hypoesthesia and/or hyperalgesia, and fMRI examinations, which included sensory stimulation of both extremities. Sensory threshold tests were found to be higher in the NP group compared with the healthy volunteers, and the heat pain threshold was found to be lower in the tests showing pain thresholds in the intergroup analyses (p < 0.05). The changes described were found in both painful and non-painful limbs. In the hypoesthetic NP group, a lower somatosensory cortex activity was found in non-painful limbs compared with the healthy volunteers (p < 0.05). In the unilateral hyperalgesic NP group, a lower somatosensory cortex activity was found on the painful side, and if the hyperalgesia was widespread, lower blood oxygen-level-dependent activity was also found in the operculum and insular cortex (p < 0.05). The patients with different phenotypes of NP had different activities in the areas related to the processing of pain, and were more prominent in patients with widespread hyperalgesia. Studies with larger numbers of patients are required for a definite statement.

Diffusion tensor imaging findings in children with sluggish cognitive tempo comorbid Attention Deficit Hyperactivity Disorder.

Objective: The construct of Sluggish Cognitive Tempo (SCT) is characterized by daydreaming, mental confusion, staring blankly and hypoactivity. Our main goal was to explore neuropsychological differences in Attention Deficit Hyperactivity Disorder-Inattentive presentation (ADHD-IA) groups with and without SCT symptoms compared to healthy controls. After detecting specific neuropsychological differences, we examined white matter microstructure using Diffusion Tensor Imaging (DTI) data obtained from 3.0 Tesla MRI scans of the cases with SCT symptoms comparing to Typically Developing (TD) controls.Method: In this study, we included 24 cases in the ADHD-IA group with SCT symptoms, 57 cases in the ADHD-IA group without SCT symptoms and, 24 children in the TD group. We applied tract-based spatial statistics to the DTI measures for obtaining fractional anisotropy (FA), axial, radial and mean diffusivity (AD, RD, MD) to explore white matter differences for the whole brain.Results: Omission error scores and longer reaction time scores were specifically associated with inattention symptoms. Commission error scores were significantly and specifically related to SCT symptoms. Cases with SCT symptoms presented increased FA in the bilateral anterior and posterior limb of the internal capsule, bilateral cerebral peduncle, and the fornix than TD group.Conclusions: Neurobiological differences in ADHD cases are still relatively unexplored. We suggest that including an assessment for SCT in the neuropsychological and neuroimaging studies of ADHD may provide more consistent results.

MRI evaluation of progressive supranuclear palsy: differentiation from Parkinson's disease and multiple system atrophy.

OBJECTIVES: To evaluate the magnetic resonance imaging (MRI)-derived parameters in differentiation of patients with progressive supranuclear palsy (PSP) from patients with Parkinson's disease (PD), multiple system atrophy (MSA), and control subjects was aimed. METHODS: Thirty-three patients [mean age, 65.21 ± 4.75 years; PSP (n = 9), MSA (n = 8), PD (n = 6), and control subjects (n = 10)] who have undergone cranial MRI were included in this retrospective study. MRI-derived parameters including areas of midbrain and pons, midbrain area-to-pons area (M/P) ratio, widths of middle cerebellar peduncle (MCP) and superior cerebellar peduncle (SCP), MCP/SCP ratio, magnetic resonance parkinsonism index (MRPI), cerebral interpeduncular angle, and length of midbrain tegmentum were compared in patients with PSP, PD, MSA, and control subjects through the analysis of variance and Kruskal-Wallis tests with Bonferroni correction and Mann-Whitney U test. RESULTS: The length of midbrain tegmentum, midbrain area, SCP, and M/P ratio were found to be lower, while cerebral interpeduncular angle and MRPI were higher in patients with PSP. Pons area, MCP width, and MCP/SCP ratio were found to be lower in patients with MSA. For PSP, cerebral interpeduncular angle has a sensitivity of 100% and specifity of 90%, and MRPI had a sensitivity of 88.9% and specifity of 100% for PSP. DISCUSSION: Several MRI-derived parameters can be used in differentiation of patients with PSP from patients with PD, MSA and control subjects. The cerebral interpeduncular angle and MRPI, which demonstrated higher values in patients with PSP, were more significant for PSP than the other parameters.

MRI assessment of pituitary iron accumulation by using pituitary-R2 in ß-thalassemia patients.

Background Patients with thalassemia major (TM) require repeated blood transfusions, which leads to accumulation of iron in a wide variety of tissues. Accumulation of iron in the pituitary gland can lead to irreversible hypogonadotropic hypogonadism (HH) in this group of patients. Purpose To investigate the reliability of pituitary-R2 as a marker to estimate the extent of pituitary iron load by comparing the pituitary magnetic resonance imaging (MRI) findings with hepatic iron load and serum ferritin levels. Material and Methods A total of 38 ß-TM patients were classified into HH (group A, n = 18) and non-HH (group B, n = 17) groups. A third group, group C, consisted of 17 healthy participants. Each participant underwent 1.5-T MRI examinations. Pituitary gland heights (PGH), pituitary-R2 values, and liver-R2 values were measured by using multi-echo spin-echo sequences. Results Pituitary-R2 values were significantly higher in group A compared with group B ( P < 0.05). A positive correlation was detected between the pituitary-R2 values and serum ferritin levels in TM patients ( P < 0.01). A threshold value of 14.1 Hz for pituitary-R2 was found to give a high specificity and sensitivity in distinguishing the TM patients with HH from those with normal pituitary functions. PGH measurements were significantly lower in group A compared with group B ( P < 0.05). Conclusion MRI-assessed pituitary-R2 seems to be a reliable marker for differentiating the TM patients with normal pituitary function from those with secondary hypogonadism due to iron toxicity.