Care during the COVID-19 pandemic hinges upon the existence of fast, safe, and highly sensitive diagnostic tools. Considering significant practical advantages of lung ultrasound (LUS) over other imaging techniques, but difficulties for doctors in pattern recognition, we aim to leverage machine learning toward guiding diagnosis from LUS. We release the largest publicly available LUS dataset for COVID-19 consisting of 202 videos from four classes (COVID-19, bacterial pneumonia, non-COVID-19 viral pneumonia and healthy controls). On this dataset, we perform an in-depth study of the value of deep learning methods for the differential diagnosis of lung pathologies. We propose a frame-based model that correctly distinguishes COVID-19 LUS videos from healthy and bacterial pneumonia data with a sensitivity of $0.90\pm 0.08$ and a specificity of $0.96\pm 0.04$. To investigate the utility of the proposed method, we employ interpretability methods for the spatio-temporal localization of pulmonary biomarkers, which are deemed useful for human-in-the-loop scenarios in a blinded study with medical experts. Aiming for robustness, we perform uncertainty estimation and demonstrate the model to recognize low-confidence situations which also improves performance. Lastly, we validated our model on an independent test dataset and report promising performance (sensitivity $0.806$, specificity $0.962$). The provided dataset facilitates the validation of related methodology in the community and the proposed framework might aid the development of a fast, accessible screening method for pulmonary diseases. Dataset and all code are publicly available at: https://github.com/BorgwardtLab/covid19_ultrasound.

(1) Background: The Culinary Medicine elective at the German medical schools of the universities of Göttingen, Giessen, and Brandenburg is a teaching kitchen-based elective aimed at training medical students on how to improve patient counselling on nutrition and lifestyle medicine topics. This curriculum was either delivered virtually (2021) or in-person (2022/2023). Changes in teaching effectiveness were evaluated. (2) Methods: The elective included seven modules in the teaching kitchen for 3 h each. It consisted of a short introduction and a hands-on interactive cooking part illustrating important dietary principles in different disease groups. The elective was conducted virtually in 2021 in a fully interactive setup using videoconference tools. Students in this cohort attended from their private kitchens whereas students in the in-person cohort (2022/2023) attended the same classes in the teaching kitchen. Standardized comparative self-assessment questionnaires on counselling competencies, nutrition knowledge, eating habits, and mental well-being (WHO-5) before and after the elective were used to determine teaching effectiveness. Paired and unpaired t-tests were performed to evaluate results. (3) Results: A total of 70 students (mean semester 6.3) were included in the virtual cohort, and 80 students (mean semester 6.3) were in the in-person cohort. In both, counselling competencies on 25 nutrition and lifestyle medicine topics increased significantly. Significant changes also occurred in most nutrition knowledge categories. Subjective well-being as well as personal attitudes towards nutrition counselling in medical practice improved significantly during the elective. Healthy eating habits improved in both groups as students ate significantly less unfavourable foods. There were no significant differences between the two groups apart from minor differences in nutrition knowledge. (4) Conclusions: The elective in Culinary Medicine improved students counselling competencies, nutrition knowledge, attitudes, well-being, and eating habits with no relevant difference between virtual and in-person teaching.

As a lightweight and non-invasive imaging technique, lung ultrasound (LUS) has gained importance for assessing lung pathologies. The use of Artificial intelligence (AI) in medical decision support systems is promising due to the time- and expertise-intensive interpretation, however, due to the poor quality of existing data used for training AI models, their usability for real-world applications remains unclear. Methods: In a prospective study, we analyze data from 63 COVID-19 suspects (33 positive) collected at Maastricht University Medical Centre. Ultrasound recordings at six body locations were acquired following the BLUE protocol and manually labeled for severity of lung involvement. Anamnesis and complete blood count (CBC) analyses were conducted. Several AI models were applied and trained for detection and severity of pulmonary infection. Results: The severity of the lung infection, as assigned by human annotators based on the LUS videos, is not significantly different between COVID-19 positive and negative patients ($p = 0.89$). Nevertheless, the predictions of image-based AI models identify a COVID-19 infection with 65% accuracy when applied zero-shot (i.e., trained on other datasets), and up to 79% with targeted training, whereas the accuracy based on human annotations is at most 65%. Multi-modal models combining images and CBC improve significantly over image-only models. Conclusion: Although our analysis generally supports the value of AI in LUS assessment, the evaluated models fall short of the performance expected from previous work. We find this is due to 1) the heterogeneity of LUS datasets, limiting the generalization ability to new data, 2) the frame-based processing of AI models ignoring video-level information, and 3) lack of work on multi-modal models that can extract the most relevant information from video-, image- and variable-based inputs. To aid future research, we publish the dataset at: https://github.com/NinaWie/COVID-BLUES.

Introduction: Although transfemoral aortic valve replacement (TAVR) is a safe treatment for elderly patients with severe aortic valve stenosis, postoperative microembolism has been described. In this secondary endpoint analysis of the POST-TAVR trial, we aimed to investigate whether changes in neuron-specific enolase (NSE)—a biomarker of neuronal damage—are associated with changes in memory function or postoperative delirium (POD). Materials and Methods: This was a prospective single-center study enrolling patients undergoing elective TAVR. Serum NSE was measured before and 24 h after TAVR. POD was diagnosed using CAM-ICU testing. Memory function was assessed before TAVR and before hospital discharge using the “Consortium to Establish a Registry for Alzheimer’s Disease” (CERAD) word list and the digit span task (DST) implemented in “∆elta-App”. Results: Subjects’ median age was 82 years (25th to 75th percentile: 77.5–85.0), 42.6% of subjects were women. CERAD scores significantly increased from pre- to post-TAVR, with p < 0.001. POD occurred in 4.4% (6/135) of subjects at median 2 days after TAVR. After TAVR, NSE increased from a median of 1.85 ng/mL (1.30–2.53) to 2.37 ng/mL (1.69–3.07), p < 0.001. The median increase in NSE was 40.4% (13.1–138.0) in patients with POD versus 17.3% (3.3–43.4) in those without POD (p = 0.17). Conclusions: Memory function improved after TAVR, likely due to learning effects, with no association to change in NSE. Patients with POD appear to have significantly higher postoperative levels of NSE compared to patients without POD after TAVR. This finding suggests that neuronal damage, as indicated by NSE elevation, may not significantly impair assessed memory function after TAVR.

Spinal metastases at the thoracolumbar junction (TLJ) pose a significant risk for spinal instability and necessitate special considerations regarding surgical management. Longer patient survival due to improved oncologic therapies may justify extensive instrumented surgery. The aim of this study was to analyze the standard of care in a large multicentric cohort of patients with TLJ metastases regarding surgical decision-making, management, and associated morbidity. Patients with surgically treated spinal metastases at the TLJ between 2010 and 2022 were enrolled at five academic tertiary spine centers. Epidemiological, surgical, clinical, and outcome data was assessed retrospectively. Surgical management was sorted according to three groups: decompression ( i ), decompression and posterior instrumentation ( ii ), and decompression and 360° instrumentation with vertebral body replacement ( iii ). Sole biopsies or kypho-/vertebroplasties were excluded. The inclusion criteria was met by 396 patients, of which 59 (15%) were treated with decompression ( i ), 235 (59%) with posterior instrumentation ( ii ), and 102 (26%) with additional vertebral body replacement ( iii ). The main factor for selection towards instrumentation was a higher SINS score (SINS 9 in ii , 10 in iii vs. 7 in i , p < 0.0001). Surgical complications occurred in 55 cases (14%), slightly more frequent following instrumentations (15% ii, iii vs. 8% i , p = 0.427). Reoperations were necessary in 65 cases (16%), mostly due to SSI (n = 19, 29%), local recurrence (n = 15, 23%), and hardware failure (HWF) during follow-up (n = 9, 18%, i-iii , p = 0.7853). HWF occurred significantly more frequent in cases with multisegmental metastases at the TLJ (p = 0.0002) which were treated with longer constructs (p = 0.0092). Median postoperative survival was 10 months. The occurrence of complications reduced postoperative survival drastically in all groups (p = 0.0023). In this large multicentric patient cohort with TLJ metastases, the majority of patients (85%) were treated with instrumented spine surgery. The main factor for patient selection towards instrumented surgery was a higher SINS score. Long instrumentations for multisegmental disease at the TLJ were identified with higher risk for hardware-failure during follow-up. In those patients, frequent follow-up imaging is warranted. As postoperative survival is drastically reduced by the occurrence of postoperative complications, it is imperative to carefully select the individually appropriate extent of surgery in order to avoid postoperative complications.

Controlling the COVID-19 pandemic largely hinges upon the existence of fast, safe, and highly-available diagnostic tools. Ultrasound, in contrast to CT or X-Ray, has many practical advantages and can serve as a globally-applicable first-line examination technique. We provide the largest publicly available lung ultrasound (US) dataset for COVID-19 consisting of 106 videos from three classes (COVID-19, bacterial pneumonia, and healthy controls); curated and approved by medical experts. On this dataset, we perform an in-depth study of the value of deep learning methods for differential diagnosis of COVID-19. We propose a frame-based convolutional neural network that correctly classifies COVID-19 US videos with a sensitivity of 0.98+-0.04 and a specificity of 0.91+-08 (frame-based sensitivity 0.93+-0.05, specificity 0.87+-0.07). We further employ class activation maps for the spatio-temporal localization of pulmonary biomarkers, which we subsequently validate for human-in-the-loop scenarios in a blindfolded study with medical experts. Aiming for scalability and robustness, we perform ablation studies comparing mobile-friendly, frame- and video-based architectures and show reliability of the best model by aleatoric and epistemic uncertainty estimates. We hope to pave the road for a community effort toward an accessible, efficient and interpretable screening method and we have started to work on a clinical validation of the proposed method. Data and code are publicly available.