Center for Artificial Intelligence in Medicine (CAIM)

Könnten digitale Pflegeasisstenten sinnvoll sein? Vier von fünf befragten Pflegefachpersonen fänden es interessant, einen solchen zu testen, berichtet Tobias Nef anlässlich der Trendtage Gesundheit in Luzern. Den vollständigen Bericht zu seinem CAIM-Projekt mit Universitäre Psychiatrische Dienste Bern (UPD) AG gibt es an unserem Forschungssymposium Ende Oktober. #AI #AIinHealthcare #KI #research #care #digitalassistant #medicine #translationalresearch

Is AI the solution to all our societal problems? Or do we need to tread carefully to maintain human supremacy and values? The reality probably lies somewhere in the middle - as Caroline Schlaufer pointed out in our CAIM Policy Talk "Narrative, Science and Policy Support". Watch the podcast now to follow the interesting discussion on the type and usage of AI narratives for policy debates, also encompassing lobbyism, narratives tainted by fake facts, and the role of universities and the media: https://lnkd.in/ega56C7D #policy #support #narrative #artificialintelligence #AI #solutions #threats #stories #political #debate #lobbies #media #science #politics #podcast #CAIMEthicsLab with KPM Center for Public Management & Multidisciplinary Center for Infectious Diseases (MCID Bern) University of Bern

Let's celebrate 2 𝒀𝒆𝒂𝒓𝒔 𝑫𝑨𝑰𝑴 - our Initiative "Diversity for AI in Medicine"! 🎇 We are proud of the 𝒄𝒐𝒎𝒎𝒖𝒏𝒊𝒕𝒚 we have been able to build with you and really enjoyed our exchange during our big March event! With now 160 members in DAIM, we are taking the opportunity to apply for the 𝑷𝒓𝒊𝒙 𝑳𝒖𝒙 of the University. So, keep your fingers crossed that we will be victorious in November! ☺️ Do you have feedback for us? What important topics should we tackle next? Do you have a tipp for an invited speaker for an online seminar or on-site scientific presentation in Bern? 👉🏼𝑻𝒉𝒆𝒏 𝒔𝒆𝒏𝒅 𝒊𝒕 𝒂𝒍𝒍 𝒂𝒍𝒐𝒏𝒈 𝒊𝒏 𝒕𝒉𝒆 𝒄𝒐𝒎𝒎𝒆𝒏𝒕𝒔! We look forward to hearing from you. 😉 #diversity #equalopportunities #fight #bias #artificialintelligence #genai #LLMs #fairness #trust #interpretability #data #curation #medicine #gender #culture #interdisciplinary #research University of Bern with Insel Gruppe Saniye Gülser Corat Enriqueta Vallejo-Yagüe Berna Özdemir Şerife Seda Kucur Ergünay Fredrick Johnson JOSEPH（乔瑟夫） Ronja Müller-Bruhn Prisca C. Quadroni - Renella Christian Lundsgaard-Hansen Stefanie Dobitz Isabelle Hahn Renu Ann Joseph, Ph.D. Johan Lundin Inti Zlobec Stavroula-Georgia Mougiakakou Mauricio Reyes Rouven Christian Porz

Can AI achieve consciousness? And if so, what terms should we negotiate with it for a prosperous co-existence? Researchers from the Universities of Bern and Amsterdam discuss how a Human-AI Deal could look like. With progress in neuromorphic architecture designs mimicking the human blueprint the differences between biological and artificial brains are increasingly vanishing. But if this is the case, we are suffering an alignment dilemma: if artificial agents share aspects of consciousness, while they (partially) overtake human intelligence, how can humans justify their own rights against growing claims of their artificial counterpart? And is AI's "being" really akin to the human condition? Check out the possible negotiated solution in this paper and tell us your thoughts in the comments! 👇🏼 https://lnkd.in/eSHdYpg6 #ArtificialIntelligence #consciousness #humans #machines #rights #deal #neuromorphic #AI #negotiate #coexistence Federico Benitez Walter Senn University of Bern Cyriel Pennartz University of Amsterdam (symbolic picture created with DALL-E)

"Patients with aggressive brain tumors unfortunately mostly have a short life expectancy. We aim to significantly reduce the time from diagnosis to treatment, currently spanning two weeks, to just two days with AI," says Ekin Ermiş, neuro-oncologist, specialized in radiotherapy for brain tumors in our latest CAIM interview. Ekin is a senior consultant at the Department of Radiation Oncology, Inselspital, University of Bern. Her current research includes the development and validation of AI tools for automatization in radiation therapy planning. She is not only convinced that "by automating the more repetitive and numerical aspects of my work I can reclaim valuable time to dedicate to patients and their families", but also appreciates the open and interdisciplinary research environment at the University of Bern and its hospital, allowing for a "can-do perception" of her task as a doctor and researcher in an increasingly digitalized surrounding. Read the interview here: https://lnkd.in/enkzc9fu #AIinhealthcare #radiation #therapy #planning #deeplearning #medicalimaging #brain #tumor #automatization #patientcare #research #innovation #biomedicalengineering #oncology Insel Gruppe University of Bern with Daniel M. Aebersold Mauricio Reyes Amith Kamath Robert Poel

Der 5. Schweizer Kongress für Telenotfallmedizin und Digital Health (13. Juni) steht dieses Jahr ganz im Zeichen von KI. Wie können wir uns die Krankenversorgung der Zukunft vorstellen? Steht die ärztliche Intuition in unvereinbarem Gegensatz zur künstlichen Intelligenz? Wie baut man KI in die medizinische Aus- und Weiterbildung ein? Wie geht man Vorurteile in der Datenlage an? Und wie empathisch kann es in einem digitalen "Smart Hospital" zugehen? Antworten auf all diese Fragen liefern sechs hochkarätige Vortragende am kostenfreien und komplett virtuellen Kongress an. Weitere Informationen und Anmeldung auf https://lnkd.in/efyMBiBX Der Kongress wird als Fortbildung anerkannt. #digitalhealth #aiinhealthcare #emergencymedicine #artificialintelligence #medicaleducation #bias #data #empathy #smarthospital #patientcare Insel Gruppe Thomas Sauter MD MME FSSERM PD Dr. med. Tanja Birrenbach, MME

Ist KI voreingenommen? In diesem "ABC"-Beitrag zu den drei Typen maschineller Lernverfahren (supervised, unsupervised or reinforcement learning) erklärt Tobias Hodel, wie KI funktioniert und welche Rolle der Mensch und die zugrundeliegenden Daten dabei spielen: https://lnkd.in/ecM7_v9r Auch spannend in dieser Reihe "Kompetenzen für die (digitale) Zukunft" ist der Beitrag "Wie diskriminieren Algorithmen?" https://lnkd.in/ezHfNDB2 ...sowie "Unser Verhältnis zu Maschinen", in dem Claus Beisbart erörtert, welche ethischen und gesellschaftlichen Herausforderungen damit verbunden wären, Maschinen (KI) als gleichberechtigte Interaktionspartner mit uns Menschen zu betrachten: https://lnkd.in/eSdKFJQq #digitalhealth #KI #AI #society #machinelearning #human #influence #manvsmachine #autonomy #bias #stereotyping #ethics #rights University of Bern #podcasts #public #education

Good news! Our last CAIM Policy Talk which was cancelled end of May is now going to be on 13 June! Join us to hear Caroline Schlaufer talk about "Narrative, Science and Policy Support"! Learn more and register for this hybrid meeting at: https://lnkd.in/eSzypWP4 #policy #healthcare #science #research #AI #aiinhealthcare #narrative #trust #regulation #digitalhealth with KPM Center for Public Management & Multidisciplinary Center for Infectious Diseases (MCID Bern) University of Bern

How do deep learning models come up with their answers? One way of testing them is Visual Question Answering – a way to probe a model’s visual understanding. Especially in medicine, this approach can help overcome the opacity of AI and build doctors’ trust in their safety for patient care. In his recently defended PhD thesis at the ARTORG Center for Biomedical Engineering Research, Sergio Tascon Morales has shown the significance of localized questions in Medical Visual Question Answering (Med-VQA) to reveal potential shortcuts or biases in DL systems for medical image analysis. "When combined with consistency, this type of question allows for a more accurate evaluation of a model's visual understanding," he explains. In this research, the use of logical relations proved to be crucial in ensuring consistency: "It enables models to recognize and avoid contradictions in their answers, much like humans do. By incorporating localized questions and logical relations, we developed a more comprehensive assessment of Med-VQA models." More about this research: https://lnkd.in/eF3YPDHb #deeplearning #VQA #medicalimaging #aiinhealthcare #transparency #XAI #consistency #bias #trust #qualityassurance #logic #machine #reasoning #visual #prompting University of Bern

𝑶𝒗𝒆𝒓𝒄𝒐𝒎𝒊𝒏𝒈 𝒓𝒐𝒕𝒂𝒕𝒊𝒐𝒏 𝒄𝒉𝒂𝒍𝒍𝒆𝒏𝒈𝒆𝒔 𝒊𝒏 𝒊𝒎𝒂𝒈𝒆 𝒔𝒆𝒈𝒎𝒆𝒏𝒕𝒂𝒕𝒊𝒐𝒏 Most segmentation networks for medical image analysis rely on standard convolutional kernels requiring expensive rotational data augmentation to learn 3D rotation equivariance. In a recently published technical paper, the Support Center for Advanced Neuroimaging (SCAN) of the Bern University Hospital has presented a segmentation Convolutional Neural Network that uses equivariant voxel convolutions based on spherical harmonics. The authors demonstrate that their approach improves segmentation performance in tasks like MRI brain tumor and healthy brain structure segmentation. The network with SO(3)-steerable convolutions is robust to data poses not seen during training and improves parameter sharing, leading to smaller network sizes and better sample efficiency. Link to paper: https://lnkd.in/dkNAhuig Code: https://lnkd.in/dp9DhK2C #CNN #medicalimaging #segmentation #3D #MRI #rotation #UNet #robustness #AIinhealthcare #neuroradiology Insel Gruppe