P09.07.A Signs and Symptoms in the last days of life of the glioblastoma patients - A Descriptive and Retrospective Study

Background

Despite scientific advances to develop techniques for early diagnosis and treatment of glioblastoma, the disease continues to have a remarkably poor prognosis. Studies have shown that in patients with glioblastoma, not only the presence of end-of-life signs and symptoms but also their frequency differs from those in the rest of cancer patients. The complexity of glioblastoma symptoms makes the integration of a palliative care philosophy essential.

Objective

To identify the main signs and symptoms that occur in patients with glioblastoma in their last seven days of life in a neuro-oncology ward.

Material and Methods

This was a descriptive and retrospective study regarding the signs and symptoms in the last seven days of life of patients with glioblastoma who were admitted to the neurology ward of a national oncology reference center. The patients were admitted after neuro-oncologist consultations between 2019 and 2020. The data collection instrument used was based on the document “The Last Hours of Living”. Confidentiality of the participants was guaranteed, and authorization was obtained from the institution’s ethics committee.

Results

The sample had 17 patients (13 men and 4 women), with an average hospital stay of 17.3 days. The men and women had average ages of 61.8 and 63 years, respectively. Clinical and nursing records revealed that the most prevalent signs and symptoms were “decreased level of consciousness” (94%); “rare and unexpected events” (94%); “respiratory dysfunction” (88%); and “loss of ability to swallow” (76%). In the last seven days of life, the approximate times of presentation for these signs and symptoms were: decreased level of consciousness, 6 days; rare and unexpected events, 2 days; respiratory dysfunction, 3 days; and loss of ability to swallow, 6 days. Support was requested from the in-hospital palliative care support team for 76.4% of patients.

Conclusion

The identification of reduced level of consciousness and alterations in swallowing as the most prevalent symptoms is consistent with the findings of previous studies. The prevalence of rare and unexpected events (such as fever, myoclonus and seizures) aligns with previous studies that indicated a difference in the signs and symptoms presented by patients with glioblastoma in the last days of life and patients with other oncological diseases. In this sense, the instrument used for data collection may not be the most appropriate for the population with primary central nervous system tumors. The decreased level of consciousness and impaired communication can complicate the assessment of signs and symptoms, such as pain, nausea, anxiety and depression. It is essential to conduct research studies with more representative samples. The recording of the signs and symptoms presented, as well as interventions and the evaluation of their effectiveness, are essential to ensure the quality of end-of-life.

Transfer learning application of a novel frequency shift convolutional neural network method for atrial fibrillation classification

Funding Acknowledgements

Type of funding sources: None.

Background

There has been a proliferation of machine learning (ML) electrocardiogram (ECG) classification algorithms reaching >85% accuracy for various cardiac pathologies. Despite the high accuracy at individual institutions, challenges remain with cross-institutional algorithm evaluation. Transfer learning (TL) is a technique in which a model trained for a specific task is repurposed for another related task, in this case ECG ML model trained at one institution used to classify ECGs at another institution. Models trained at one institution, however, might not be generalizable for accurate classification when deployed broadly due to differences in type, time, and sampling rate of traditional ECG acquisition. TL would thus require additional ECG signal processing which inherently introduces noise and has high computational costs. Our previous work has shown that frequency domain (FD) convolutional neural networks (CNN) outperform traditional time domain (TD) CNN methods in detecting atrial fibrillation (Afib) and are more robust against noise and sampling variations.

Purpose

In this study, we explore ECG transfer learning by comparing performance of TD and FD CNN ECG classification models. Examine the hypothesis that FD signal analysis can overcome limitations inherent to TD and demonstrate application for Afib classification in three different publicly available datasets.

Methods

PTB-XL ECG dataset was used to train TD and FD CNN models for Afib classification. TL on the two data sets, Lobachevsky University Electrocardiography Database (LUDB) and KURIAS-ECG database (KURIAS), was performed with portions of each dataset added to PTB-XL training data before test runs on the remaining ECGs. The models were also tested directly on the two datasets (cross-testing) for comparison.

Results

Results of TL are summarized in Fig. 1. Panels (a) and (b) show TD and FD performance on PTB-XL dataset, with FD outperforming TD. Panels (c) and (d) show TL of PTB-XL trained TD and FD models applied on LUDB Afib data with FD outperforming TD. Panels (e) and (f) show TL of PTB-XL trained models applied to KURIAS Afib ECGs, with FD again outperforming TD. Table 1 summarizes results of TL and direct cross-testing with TL outperforming cross-testing.

Conclusions

FD models were superior to TD models in Afib classification, both in cross-testing and TL. TL technique, in which a pre-trained model is used as starting point for novel dataset training, outperformed direct cross-testing. TL with FD CNN has the capacity to leverage models trained on a particular set of ECGs to classify ECGs with varying signal acquisition properties. FD CNN may be robust to inter-institutional variability and has potential for widespread application with no compromise to ECG classification power. A potential application of these findings is deployment of ECG classification CNN models trained at institutions with large ECG databases to hospitals with smaller datasets.

135 RECOMBINANT BOVINE SOMATOTROPIN ON THE IN VITRO PRODUCTION OF BOVINE EMBRYOS

Bovine growth hormone (bGH) has been used to improve the results for in vitro production of bovine embryos. Inclusion of bGH in the maturation medium increases both rate of cleavage and frequency of blastocyst development. Thus, the purpose of the present study was to evaluate the effect of recombinant bovine somatotropin (rBST) on cleavage and blastocyst development of bovine embryos when included during in vitro maturation (IVM) only (Group 1), during both IVM and in vitro culture (IVC; Group 2), during IVC only (Group 3), or not included during either IVM or IVC (Group 4). Specifically, in Group 1, oocytes were matured in TCM 199 (Earle's salts) supplemented with 10% FCS, LH, FSH, oestradiol, and amikacin (IVM medium), plus 100 ng mL–1 of rBST and cultured in SOFaaci supplemented with essential amino acids, tri-sodium citrate, myo-inositol, and 5% FBS. In group 2, oocytes were matured in IVM medium containing 100 ng mL–1 of rBST and cultured in SOFaaci supplemented with essential amino acid, tri-sodium citrate, myo-inositol, 5% FBS; on Day 5, rBST (50 ng mL–1) was added. In Group 3, oocytes were matured in IVM medium without rBST; on Day 5, rBST (50 ng mL–1) was added. Group 4 was the control, without rBST supplementation. The treatment groups were analysed using the SAS® (SAS Institute Inc., Cary, NC, USA) in a completely randomised design (P < 0.05). Somatotropin has receptors in cumulus cells and in the zona pellucida acting directly in the oocyte; however, the increase in cleavage rate seen in previous studies after rBST treatment was not observed in the present study. Supplementation of culture medium with rBST during Day 2 to 6 of IVC has been shown to increase the number of trophoblast and subsequent pregnancy rate following transfer. However, in the present study, addition of 50 and 100 ng mL–1 of rBST to the maturation or culture medium did not affect the cleavage rate of embryos and blastocyst production. Table 1.Analysis of the meaning and percentages related to cleavage and production of embryos

Anti-inflammatory effect of bioflora probiotic administered orally or subcutaneously with live or dead bacteria

The anti-inflammatory effect of Bioflora probiotic administered orally or subcutaneously with viable or nonviable bacteria was analyzed in two experimental models of randomly selected female Wistar rats. The use of indomethacin at a dose of 50 mg/kg was associated with gastric necrotic lesions and multiple erosions of the small intestine, with marked mucosal neutrophil infiltrate measured by myeloperoxidase (MPO). Probiotics prevented both gastrointestinal lesions and neutrophil infiltrate (p < 0.001). In the model of carrageenin-induced plantar edema in the rat, the oral or subcutaneous administration of Bioflora with live or dead bacteria proved to prevent plantar edema (p < 0.001). We concluded that Bioflora probiotic given orally or subcutaneously with live or dead bacteria has an anti-inflammatory effect.

Substance P facilitates memory in goldfish in an appetitively motivated learning task

The aim of this study was to investigate if the neuropeptide substance P (SP) can improve learning and memory in goldfish (Carassius auratus). Four groups of fish were trained for seven days to find food in one out of two compartments until discrimination was achieved. On the last training day, they were injected (intra-abdominal) with haloperidol or vehicle before the training, and with SP or vehicle immediately after the training session. Each group of fish received either: (1) vehicle+vehicle (n = 18); (2) vehicle + SP, (n = 20, SP 50 mg/kg); (3) haloperidol+ vehicle (n = 15, haloperidol 2 mg/kg); or (4) haloperidol+ SP (n = 14, haloperidol 2 mg/kg, SP 50 mg/kg). Twenty-four hours later, the time spent to find the food was recorded. Reversal training was done for four consecutive days after this post-injection test and the time spent to find the food was recorded again. The results indicate that only the group treated with vehicle + SP needed more time to reach reversal training than control fish (Mann-Whitney U-test, P = 0.0009). It is suggested that SP can enhance memory in fish and that this effect may have a dopaminergic mediation in discrimination learning task.

Reinforcing properties of the neuropeptide substance P in Carassius auratus: evidence of dopaminergic system involvement

The aim of the present study was to investigate whether neuropeptide substance P (SP) has reinforcing effects in Carassius auratus and whether this effect could be related to dopaminergic neurotransmission. For this purpose fishes were put in a three-compartment box in which one compartment gave access to two others that did not directly link. The time spent in each compartment was registered for 10 min to determine a possible preferred compartment. Twenty-four hours later, the fish were given one of the following intraperitoneal treatments: a) Group VEH, injected with the vehicle of substance P; b) Group SP25, injected with SP, 25 micrograms/kg body wt.; c) Group SP50, injected with SP, 50 micrograms/kg; or d) Group HALO, injected with haloperidol (2 mg/kg) 30 min before an injection of SP (50 micrograms/kg). Immediately after the treatment the fish were kept for 30 min in the compartment preferred least the day before. On the next day the fish were retested for 10 min to verify the time spent in each compartment. The results indicate that SP at the dose of 50 micrograms/kg enhanced the time spent on the paired compartment, and that the pretreatment with haloperidol abolished this enhancement. It is suggested that SP has reinforcing effects in C. auratus that may be mediated by the dopaminergic system.