Spatial transcriptomics (ST) methods unlock molecular mechanisms underlying tissue development, homeostasis, or disease. However, there is a need for easy-to-use, high-resolution, cost-efficient, and 3D-scalable methods. Here, we report Open-ST, a sequencing-based, open-source experimental and computational resource to address these challenges and to study the molecular organization of tissues in 2D and 3D. In mouse brain, Open-ST captured transcripts at subcellular resolution and reconstructed cell types. In primary head-and-neck tumors and patient-matched healthy/metastatic lymph nodes, Open-ST captured the diversity of immune, stromal, and tumor populations in space, validated by imaging-based ST. Distinct cell states were organized around cell-cell communication hotspots in the tumor but not the metastasis. Strikingly, the 3D reconstruction and multimodal analysis of the metastatic lymph node revealed spatially contiguous structures not visible in 2D and potential biomarkers precisely at the 3D tumor/lymph node boundary. All protocols and software are available at https://rajewsky-lab.github.io/openst.

In September 2021 I made a collection of interview transcripts available for public use under a CreativeCommons license through the Princeton DataSpace. The interviews include 39 conversations I had with gig workers at AmazonFlex, Uber, and Lyft in 2019 as part of a study on automation efforts within these organizations. I made this decision because (1) I was required to contribute to a publicly available data set as a requirement of my funding and (2) I saw it as an opportunity to engage in the collaborative qualitative science experiments emerging in Science and Technology studies. This article documents my thought process and step-by-step design decisions for designing a study, gathering data, masking it, and publishing it in a public archive. Importantly, once I decided to publish these data, I determined that each choice about how the study would be designed and implemented had to be assessed for risk to the interviewee in a very deliberate way. It is not meant to be comprehensive and cover every possible condition a researcher may face while producing qualitative data. I aimed to be transparent both in my interview data and the process it took to gather and publish these data. I use this article to illustrate my thought process as I made each design decision for this study in hopes that it could be useful to a future researcher considering their own data publishing process.

BACKGROUND: The accuracy of movement determination software in current activity trackers is insufficient for scientific applications, which are also not open-source.
OBJECTIVE: To address this issue, we developed an accurate, trainable, and open-source smartphone-based activity-tracking toolbox that consists of an Android app (HumanActivityRecorder) and 2 different deep learning algorithms that can be adapted to new behaviors.
METHODS: We employed a semisupervised deep learning approach to identify the different classes of activity based on accelerometry and gyroscope data, using both our own data and open competition data.
RESULTS: Our approach is robust against variation in sampling rate and sensor dimensional input and achieved an accuracy of around 87% in classifying 6 different behaviors on both our own recorded data and the MotionSense data. However, if the dimension-adaptive neural architecture model is tested on our own data, the accuracy drops to 26%, which demonstrates the superiority of our algorithm, which performs at 63% on the MotionSense data used to train the dimension-adaptive neural architecture model.
CONCLUSIONS: HumanActivityRecorder is a versatile, retrainable, open-source, and accurate toolbox that is continually tested on new data. This enables researchers to adapt to the behavior being measured and achieve repeatability in scientific studies.

Characterizing neurons by their electrophysiological phenotypes is essential for understanding the neural basis of behavioral and cognitive functions. Technological developments have enabled the collection of hundreds of neural recordings; this calls for new tools capable of performing feature extraction efficiently. To address the urgent need for a powerful and accessible tool, we developed ElecFeX, an open-source MATLAB-based toolbox that (1) has an intuitive graphical user interface, (2) provides customizable measurements for a wide range of electrophysiological features, (3) processes large-size datasets effortlessly via batch analysis, and (4) yields formatted output for further analysis. We implemented ElecFeX on a diverse set of neural recordings; demonstrated its functionality, versatility, and efficiency in capturing electrical features; and established its significance in distinguishing neuronal subgroups across brain regions and species. ElecFeX is thus presented as a user-friendly toolbox to benefit the neuroscience community by minimizing the time required for extracting features from their electrophysiological datasets.

BACKGROUND: In 2023, the release of ChatGPT triggered an artificial intelligence (AI) boom. The underlying large language models (LLM) of the nonprofit organization \"OpenAI\" are not freely available under open-source licenses, which does not allow on-site implementation inside secure clinic networks. However, efforts are being made by open-source communities, start-ups and large tech companies to democratize the use of LLMs. This opens up the possibility of using LLMs in a data protection-compliant manner and even adapting them to our own data.
OBJECTIVE: This paper aims to explain the potential of privacy-compliant local LLMs for radiology and to provide insights into the \"open\" versus \"closed\" dynamics of the currently rapidly developing field of AI.
METHODS: PubMed search for radiology articles with LLMs and subjective selection of references in the sense of a narrative key topic article.
RESULTS: Various stakeholders, including large tech companies such as Meta, Google and X, but also European start-ups such as Mistral AI, contribute to the democratization of LLMs by publishing the models (open weights) or by publishing the model and source code (open source). Their performance is lower than current \"closed\" LLMs, such as GPT‑4 from OpenAI.
CONCLUSIONS: Despite differences in performance, open and thus locally implementable LLMs show great promise for improving the efficiency and quality of diagnostic reporting as well as interaction with patients and enable retrospective extraction of diagnostic information for secondary use of clinical free-text databases for research, teaching or clinical application.
UNASSIGNED: HINTERGRUND: Die Veröffentlichung von ChatGPT löste im Jahr 2023 einen KI-Boom aus. Die zugrundeliegenden großen Sprachmodelle (Large Language Models, LLM) der gemeinnützigen Organisation „OpenAI“ sind nicht frei unter Open-Source-Lizenzen verfügbar, was keine Implementierung vor Ort in gesicherten Kliniknetzen erlaubt. Von Open-Source-Gemeinschaften, Start-ups, aber auch von großen Tech-Firmen gibt es jedoch Bestrebungen, die Anwendung von LLMs zu demokratisieren. Dies bietet die Möglichkeit, LLMs auch datenschutzkonform anzuwenden und sogar auf eigene Daten anzupassen.
UNASSIGNED: In diesem Beitrag soll das Potenzial von datenschutzkonformen, lokalen LLMs für die Radiologie erläutert und Einblicke in die „open“ versus „closed“ Dynamik der aktuell rasanten Entwicklungen im Bereich künstlicher Intelligenz (KI) gegeben werden.
METHODS: PubMed-Suche zu radiologischen Artikeln mit LLMs und subjektive Auswahl von Referenzen im Sinne eines narrativen Leitthemenartikels.
UNASSIGNED: Verschiedene Akteure, darunter große Tech-Firmen wie Meta, Google und X, aber auch europäische Start-ups wie Mistral AI, tragen zur Demokratisierung von LLMs durch Veröffentlichung der Modelles („open weights“) oder durch Veröffentlichung von Modell und Quellcode („open source“) bei. Ihre Performanz ist geringer als aktuelle „closed“ LLMs, wie z. B. GPT‑4 von OpenAI.
CONCLUSIONS: Trotz Unterschieden in der Performanz zeigen offene und damit lokal implementierbare LLMs großes Potenzial zur Verbesserung der Effizienz und Qualität der Befundung sowie der Interaktion mit Patienten und ermöglichen eine retrospektive Extraktion von Befundinformationen zur Sekundärnutzung klinischer Freitextdatenbanken für Forschung, Lehre oder klinische Anwendung.

The OpenFlexure Microscope is an accessible, three-dimensional-printed robotic microscope, with sufficient image quality to resolve diagnostic features including parasites and cancerous cells. As access to lab-grade microscopes is a major challenge in global healthcare, the OpenFlexure Microscope has been developed to be manufactured, maintained and used in remote environments, supporting point-of-care diagnosis. The steps taken in transforming the hardware and software from an academic prototype towards an accepted medical device include addressing technical and social challenges, and are key for any innovation targeting improved effectiveness in low-resource healthcare. This article is part of the Theo Murphy meeting issue \'Open, reproducible hardware for microscopy\'.

UNASSIGNED: At the intersection of neural monitoring and decoding, event-related potential (ERP) based on electroencephalography (EEG) has opened a window into intrinsic brain function. The stability of ERP makes it frequently employed in the field of neuroscience. However, project-specific custom code, tracking of user-defined parameters, and the large diversity of commercial tools have limited clinical application.
UNASSIGNED: We introduce an open-source, user-friendly, and reproducible MATLAB toolbox named EPAT that includes a variety of algorithms for EEG data preprocessing. It provides EEGLAB-based template pipelines for advanced multi-processing of EEG, magnetoencephalography, and polysomnogram data. Participants evaluated EEGLAB and EPAT across 14 indicators, with satisfaction ratings analyzed using the Wilcoxon signed-rank test or paired t-test based on distribution normality.
UNASSIGNED: EPAT eases EEG signal browsing and preprocessing, EEG power spectrum analysis, independent component analysis, time-frequency analysis, ERP waveform drawing, and topological analysis of scalp voltage. A user-friendly graphical user interface allows clinicians and researchers with no programming background to use EPAT.
UNASSIGNED: This article describes the architecture, functionalities, and workflow of the toolbox. The release of EPAT will help advance EEG methodology and its application to clinical translational studies.

Medical research uses increasingly massive, complex and interdependent data, the analysis of which requires the use of specialized algorithms. In order to independently reproduce and validate the results of a scientific study, it is no longer sufficient to share the text of the article as an open-access document, together with the raw research data according to the open-data approach. It is now also needed to share the algorithms used to analyze the data with other research teams. Free and open-source software precisely responds to this need to disseminate technical knowledge at a large scale. In this paper, we present several examples of free software used in medicine, with a particular focus on medical imaging.
La recherche médicale recourt à des données de plus en plus massives, complexes et interdépendantes dont l’analyse nécessite l’usage d’algorithmes spécialisés. Afin de reproduire et valider les résultats d’une étude scientifique de manière indépendante, il n’est, dès lors, plus suffisant de partager le texte de l’article en «open-access» complété avec les données brutes en «open-data». Il convient désormais d’également partager les algorithmes qui ont servi à l’analyse des données avec d’autres équipes de chercheurs. Le logiciel libre et «open-source» répond précisément à ce besoin de diffuser les connaissances techniques à grande échelle. Dans cet article, nous présentons plusieurs exemples de logiciels libres utilisés en médecine, avec une attention particulière portée à l’imagerie médicale.

Access to medical technologies is a critical component of universal access to care; however, the advancement of technologies for children has historically lagged behind those for adults. The small market size, anatomic and physiologic variability, and legal and ethical implications pose unique barriers to developing and commercialising paediatric biomedical innovations. These challenges are magnified in low-resource settings (LRS), which often lack appropriate regulatory oversight, support for service contracts, and supply chain capacity. The COVID-19 pandemic exposed shortcomings in the traditional industry model for medical technologies, while also catalysing open-source approaches to technology development and dissemination. Open-source pathways - where products are freely licenced to be distributed and modified - addressed key shortages in critical equipment. Relatedly, we argue that open-source approaches can accelerate paediatric global health technology development. Open-source approaches can be tailored to clinical challenges independent of economic factors, embrace low-cost manufacturing techniques, and can be highly customisable. Furthermore, diverse stakeholders, including families and patients, are empowered to participate in collaborative communities of practice. How to regulate the development, manufacture, and distribution of open-source technologies remains an ongoing area of exploration. The need for democratised innovation must be carefully balanced against the imperatives of safety and quality for paediatric-specific solutions. This can be achieved, in part, through close coordination between national regulatory agencies and decentralised networks where products can be peer-reviewed and tested. Altogether, there is significant potential for open source to advance more equitable and sustainable medical innovations for all children.

Purpose To assess the performance of a local open-source large language model (LLM) in various information extraction tasks from real-life emergency brain MRI reports. Materials and Methods All consecutive emergency brain MRI reports written in 2022 from a French quaternary center were retrospectively reviewed. Two radiologists identified MRI scans that were performed in the emergency department for headaches. Four radiologists scored the reports\' conclusions as either normal or abnormal. Abnormalities were labeled as either headache-causing or incidental. Vicuna (LMSYS Org), an open-source LLM, performed the same tasks. Vicuna\'s performance metrics were evaluated using the radiologists\' consensus as the reference standard. Results Among the 2398 reports during the study period, radiologists identified 595 that included headaches in the indication (median age of patients, 35 years [IQR, 26-51 years]; 68% [403 of 595] women). A positive finding was reported in 227 of 595 (38%) cases, 136 of which could explain the headache. The LLM had a sensitivity of 98.0% (95% CI: 96.5, 99.0) and specificity of 99.3% (95% CI: 98.8, 99.7) for detecting the presence of headache in the clinical context, a sensitivity of 99.4% (95% CI: 98.3, 99.9) and specificity of 98.6% (95% CI: 92.2, 100.0) for the use of contrast medium injection, a sensitivity of 96.0% (95% CI: 92.5, 98.2) and specificity of 98.9% (95% CI: 97.2, 99.7) for study categorization as either normal or abnormal, and a sensitivity of 88.2% (95% CI: 81.6, 93.1) and specificity of 73% (95% CI: 62, 81) for causal inference between MRI findings and headache. Conclusion An open-source LLM was able to extract information from free-text radiology reports with excellent accuracy without requiring further training. Keywords: Large Language Model (LLM), Generative Pretrained Transformers (GPT), Open Source, Information Extraction, Report, Brain, MRI Supplemental material is available for this article. Published under a CC BY 4.0 license. See also the commentary by Akinci D\'Antonoli and Bluethgen in this issue.