Patrick Malone, MD PhD

🧠 ⚙ #𝐍𝐄𝐔𝐑𝐎𝐓𝐄𝐂𝐇 𝐍𝐎𝐓𝐀𝐁𝐋𝐄𝐒 #8: 𝐌𝐚𝐲 𝟏-𝟏𝟓, 𝟐𝟎𝟐𝟒 This month's part 1 roundup covers a (relatively) slow fortnight of #neurotechnology developments while I've been working remotely in #washingtonDC As usual, there's been no shortage of #BCI news, with important headlines from Neuralink and Synchron, while Integral Neurotechnologies emerged from stealth led by Milan Cvitkovic. I also weigh in on the trust gap in neurotech, pointing to an excellent (and damning) STAT investigation about a neurosurgeon behaving badly at Mt. Sinai under the guise of #deepbrainstimulation research. As the field of #braincomputerinterface races to hit commercial milestones, there's still important work left to do... 🗞 Other market news of note: Fresh funding, New partnerships, and a SPAC. 🤔 You'll also find a link to a members-only post: 𝙰𝚟𝚘𝚒𝚍𝚒𝚗𝚐 𝙽𝚎𝚞𝚛𝚘𝚝𝚎𝚌𝚑'𝚜 𝙽𝚎𝚡𝚝 𝚃𝚑𝚎𝚛𝚊𝚗𝚘𝚜 🗓 New Events posted for May and June. 💼 And - a brand new feature: Introducing ＪＯＢ ＰＯＳＴＩＮＧＳ! Are you a cool company hiring for cool roles in neurotechnology? Email/DM me with your postings. Please note that a small placement fee applies. More exciting announcements for subscribers in the coming weeks...it's a good time to Subscribe! As usual, I'm asking for your help to spread the word, share with your colleagues, teams, and professional networks. It really makes a difference. Thanks! Finally a SPECIAL SHOUT OUT to the two executives who paused their busy schedules to share candid, critical feedback on my analysis, not to get a soundbyte in, but to make sure I am understanding the nuances of this market and its key players. I'm here to learn... 🙏 Thanks so much for reading and engaging. I'd love to hear from you in the comments, DM, email, text... LET'S GO. https://lnkd.in/edxcbcMb

46

🧠 Zeto / Rapid Full-Montage EEG gets FDA 510(k) approval for its next-gen EEG solution, ONE The new product, dubbed ONE, is a portable EEG device that enables bedside seizure detection, along with video, audio, and live remote monitoring and interpretation. It includes a user-friendly headset equipped with 21 soft-tip electrodes, which are positioned as per the commonly followed 10-20 EEG system. The system also features an intuitive LED feedback mechanism for easy adjustments and requires minimal training for any healthcare professional to operate effectively. Zeto / Rapid Full-Montage EEG founder and CEO Aswin Gunasekar said: “ONE is a culmination of 8+ years of customer-centric product development by our industrious team. “As our first-generation Wrap (WR19) headset sees increasing adoption for short-term EEGs, ONE strengthens our portfolio by democratizing access to continuous EEG brain monitoring and diagnostics. “We anticipate ONE to notably advance EEG monitoring in critical care, outpatient and home settings.” Read more online: https://lnkd.in/eYtABMqT 📰 Follow Medical Device Developments to receive the latest medical device news daily and to subscribe to our weekly newsletter #MedicalDeviceDevelopments #medicaldevices #medicalmanufacturing

54

5 Comments

Neural Radiance Fields in Medical Imaging: A Brief Summary 💫

8

Join us on June 5, 2024, at 1 PM (ET) for a webinar on tackling the reproducibility and generalizability challenges in #MachineLearning ! Learn strategies for data accessibility, leakage prevention, preprocessing, algorithm selection, and more to ensure reliable ML outcomes. Farhad Maleki Society for Imaging Informatics in Medicine (SIIM) Register: ⬇ https://lnkd.in/dqrDUg2J Objectives: - Equip participants with strategies to facilitate data accessibility, prevent data leakage, and standardize data preprocessing to enhance the reproducibility of machine learning projects. - Provide insights into selecting appropriate algorithms, promoting code sharing, and managing dependencies effectively to achieve reliable machine learning outcomes. - Advocate for the adoption of comprehensive and transparent research documentation standards to improve the clarity and reproducibility of machine learning research. - Demonstrate how to utilize advanced tools and frameworks and adhere to community standards to significantly improve the reproducibility of machine learning systems. #ML #DataScience #Healthcare #Webinar #AI

14

neuroClues lands €5m for eye-tracking tech that detects #neurodegenerative disease https://lnkd.in/g7VSpmbi Startup aims to diagnose multiple neurological disorders ‘years before the manifestation of clinical symptoms.’ French-Belgian medtech company NeuroClues has secured €5 million in new investment to support the launch of its technology designed to support the early diagnosis of neurological disorders. The firm has its initial sights set on the early detection of Parkinson’s disease, with other neurodegenerative conditions also in its roadmap, including Alzheimer’s and multiple sclerosis. With the vision of becoming the “stethoscope of the brain,” NeuroClues has developed an eye-tracking technology that eliminates the need for calibration and promises earlier diagnosis compared to conventional methods. Leveraging a portable and connected headset, the system captures up to 800 infrared images per eye per second, providing healthcare practitioners with objective and quantifiable biomarkers for various neurological disorders. NeuroClue's technology has been integrated into the Iceberg study at the Brain Institute within La Salpêtrière Hospital in Paris, which aims to advance research on Parkinson’s disease, that aims to identify biomarkers for the early detection and monitoring of Parkinson’s disease progression. “Subtle alterations in eye movement parameters can precede the onset of gait and balance disorders, allowing for the early detection of brain circuit dysfunctions in Parkinson’s disease before symptoms appear,” said Brain Institute professor Marie Vidailhet. “This early detection will enable the implementation of earlier care by strengthening the patient’s abilities. This can involve physical and sports training, monitoring, and management of risk factors: the patient becomes their own therapist!” Omega Human Aude Monciino Toni Masselter

2

🚀 Digital Twins from Synthetic Data for Clinical Trials Attract OpenAI 🚀 The era of Digital Twins has finally arrived. Introduced in my 2018 book 'Your Guide to Delight,' the number of relevant publications starts to increase, first true applications become more mature, and novel creative applications emerge. See, for instance, my post on Saturday (May 10) on a sophisticated digital twin model that simulates real-life alcohol consumption and combines #biomarkers with a #digitalhealth app. On the eve of the 1st International Symposium on Digital Twins in Healthcare, here is some novel exciting news. Unlearn.AI, which has built a machine learning platform that creates "digital twin" profiles of patients in clinical trials, is expanding its board with a significant profile and achieving significant regulatory and business milestones. The company won approval from the European Medicines Agency (EMA) for using its AI-driven approach for conducting smaller and quicker clinical trials and struck several multi-million dollar deals with pharmaceutical companies, including Merck. Mira Murati, chief technology officer at OpenAI (from ChatGPT fame), is joining the startup's board. "The team at Unlearn is working on applications of AI that have incredible potential to revolutionize healthcare, diagnostics, and treatment." While #generativeAI is so far known for creating content like text, images, video, and even computer code, Unlearn uses it to build #digitaltwins to speed up clinical drug trials through its neural networks. They create so-called synthetic (patient) data. If successful, Unlearn's software could lower costs for drug developers by cutting the number of enrollments needed for clinical trials by replacing patients who receive a placebo. 🗣️ I will discuss these groundbreaking developments in my closing keynote at the 1st International Symposium on Digital Twins in Healthcare in Cyprus this week. Learn More Here: https://lnkd.in/eamYU4Ec 🎥 Dive deeper into how AI is revolutionizing healthcare by visiting my YouTube channel Koen Kas YouTube and explore more about our latest innovations and insights : https://lnkd.in/d47khtbg 🌐 For more details on my keynotes like "AI in Healthcare: Demystified and Destigmatised" and "How AI Will Propel Us from Healthy & Happy to Being Epic," visit my website https://lnkd.in/d2Quv5w3 📚 Also, don't miss out on pre-ordering my latest book "EPIC: Unleash your full potential through health, art, and technology": Dutch Edition Pre-Order: https://lnkd.in/d_Kbwph9 English Edition Pre-Order: https://lnkd.in/dxZ243ja Kudos to the Conference Chairs Adrian M. Ionescu and Andreani Odysseos and a special shoutout to Charles Fisher and his team for their groundbreaking work. #syntheticdata #digitaltwin #drugdiscovery #genAI #drugdevelopment #digitalhealth #pharma

110

6 Comments

“These approaches could be the enabler of robust digital twins of individual patients (i.e., representations of up-to-date individual patient data in digital form, serving as a record of patient health and enabling personalized predictive analytics) with LLMs used to rapidly create stable individual patient KGs (knowledge graphs) as stable robust data structures, which could be used to augment and verify data interpreted by LLMs from newly conducted consultations.”

9

1 Comment

☑️ *READ ABSTRACT BELOW:* We address the interpretability of the machine learning algorithm in the context of the relevant problem of discriminating between patients with major depressive disorder (MDD) and healthy controls using functional networks derived from resting-state functional magnetic resonance imaging data. We applied linear discriminant analysis (LDA) to the data from 35 MDD patients and 50 healthy controls to discriminate between the two groups utilizing functional networks' global measures as the features. We proposed the combined approach for feature selection based on statistical methods and the wrapper-type algorithm. This approach revealed that the groups are indistinguishable in the univariate feature space but become distinguishable in a three-dimensional feature space formed by the identified most important features: mean node strength, clustering coefficient, and the number of edges. LDA achieves the highest accuracy when considering the network with all connections or only the strongest ones. Our approach allowed us to analyze the separability of classes in the multidimensional feature space, which is critical for interpreting the results of machine learning models. We demonstrated that the parametric planes of the control and MDD groups rotate in the feature space with increasing the thresholding parameter and that their intersection increases with approaching the threshold of 0.45, for which classification accuracy is minimal. Overall, the combined approach for feature selection provides an effective and interpretable scenario for discriminating between MDD patients and healthy controls using measures of functional connectivity networks. This approach can be applied to other machine learning tasks to achieve high accuracy while ensuring the interpretability of the results. Andreev AV, Chaos. 2023 Jun 1;33(6):063140. doi: 10.1063/5.0155567. PMID: 37318340. #Gesundheit #Bildung #Fuehrung #Coaching #Mindset #Motivation #Gehirn #Neuroscience #Psychologie #Persoenlichkeitsentwicklung #Kindheit #KeyNoteSpeaker #Humangenetik #Biochemie #Neuroleadership #Ernaehrung #Transformation #Stress #Demografie #Gender #Age #interkulturelleKompetenz #Epigenetik #Veraenderung #EmotionaleIntelligenz #Change #Gesellschaft #Organisationsentwicklung #Philosophie #Beratung # Quantum

1

There is currently an arms race unfolding in front of our eyes, and the stakes are high. Who will lead the way? With Meta releasing LLAMA (a open source model for research), Microsoft pouring billions into OpenAI (creator of ChatGPT), and Google developing GEMINI (a task-oriented conversational AI), the tech giants are fighting for dominance. Beyond LLMs they are then looking to produce the 'currency' of the future. Microchips. They don't want to depend on players like NVIDIA to provide the brains behind their AI models. I believe, as many others do, whoever holds the fastest microchip holds an unbelievable amount of power in the near future. And this isn't just about better search results – AI is ready to revolutionize industries like healthcare (more accurate diagnoses), customer service (hyper-personalized chatbots), and content creation (AI-assisted writing tools). Sometimes I question whether the AI powered future is in the right hands – those of profit-focused corporations whose actions can leave extreme consequences for the world. How do we balance innovation with responsible development that benefits all of humanity?

2

The important milestone of medical AI development was highlighted by Eric Topol, MD in his Ground Truth publication "When Medical A.I. is Lifesaving. Counterbalanced by some disappointments". At the end of April, research group from Taiwan published a paper describing results of a  multisite randomized controlled trial evaluating the ability of an AI-enabled electrocardiogram analysis tool to identify hospitalized patients with a high risk of mortality. The study was appeared in Nature Medicine and included 39 physicians and 15,965 patients. Electrocardiograms of patients in interventional arm were analyzed with AI and physicians received alert with report and warning messages, flagging patients predicted to be at high risk of mortality. The control arm included patients whose ECGs were analyzed by conventional method (by a physician with no AI involvement). The implementation of the AI-ECG alert was associated with a significant reduction (17%) of all-cause mortality within 90 days and saving 7 lives per 100 in some groups. As Dr. Topol pointed out, the results are comparable to most successful medical treatments, such as statins. Notably, it was the first-ever RCT to estimate the effect of AI-based intervention on patients' mortality. However, the results of four trials of real world scenarios of Generative AI use, as described in the same Ground Truth publication, were not so promising.  Large Language Models (including Chat GPT 4) were unsuccessful in answering oncology questions and assisting in medical coding. They also failed to reduce clinicians’ time spent writing reply messages to patients (though they did generate more empathetic replies) and posed a risk of severe harm in 7% of generated messages which is definitely not acceptable. Surely, as Dr. Topol stated “Progress in medical A.I. won’t occur in a straight line”. It has been repeatedly confirmed that AI is far better than humans at analyzing repeated patterns not generated by humans, such as MRI or ECG. However, when it comes to working with more abstract entities, such as diagnoses or clinical messages, its effectiveness is much more modest. In any case, I strongly encourage you to refer to the original Ground Truth publication for more details - https://lnkd.in/dXaEAq6m

21

2 Comments

Infographic: Visual Analysis of the Beatles I built a complex alt text generator for large, information-rich visuals by request. While testing, I've processed thousands of vintage maps, schematics, and infographics. They're paired with a visual description but very cool in their own right! Tons of curriculum applications, among other things. Get ahold of me if you'd like to try the AI! (or join https://buff.ly/4bZiXvf for automatic access)

1

How is your neural network doing? I don't mean the one in your computer, I mean the one inside your head 😀 A new article in Nature Medicine (link below) demonstrates that neural networks (in your actual biological brain) show specific electrical signals which indicate clinical states such a depression and anxiety. What does this mean? First - it is duh! - of course. The main activity of brain is passing electric current between cells so its state (of function and dysfunction) can be computed by analysing patterns of electric conduction should not come as a surprise. But in reality, it is more interesting because the field of #neuropsychiatry is struggling to find measurable, repeatable, reliable signal beyond 'crude' behaviour signals to diagnose, monitor and treat these debilitating conditions which are a growing burden on society. And this may be a helpful approach. This research could pave the way for more personalized treatments. Imagine a future where your treatment plan is tailored to your brain's unique electrical patterns, potentially increasing the effectiveness of interventions and reducing side effects. Moreover, this approach could help in early detection. By identifying specific electrical patterns associated with depression and anxiety, clinicians might be able to intervene before these conditions become severe, improving patient outcomes. Currently, 30–40% of patients do not achieving remission after trying one treatment Finally, neurostimulation techniques, such as transcranial magnetic stimulation (#TMS) and deep brain stimulation (#DBS), could leverage these findings to target specific brain circuits associated with depression and anxiety, offering new hope for patients who do not respond to traditional treatments. Additionally, neural interfaces, like brain-computer interfaces (#BCIs), could be developed to monitor and modulate brain activity in real-time, providing continuous feedback and personalized adjustments to treatment protocols, potentially enhancing therapeutic outcomes. The UK's Advanced Research + Invention Agency (ARIA) is also exploring these possibilities. Their program aims to develop new tools that can interact with the central nervous system at the circuit level to understand and treat these conditions. Check out their call for neural interfaces and follow Jacques Carolan for updates.

25

3 Comments

Recent changes to mlx-tuning-fork (and MLX) allow me to use Llama3's prompt format when prompting a model (and its adapters and checkpoints) in training. As it is training (a 3-4 day process even for a 7B model), I can prompt its checkpoints, which is useful for a new version of MrGrammaticaOntology I am working on. It is trained on medqa_corpus_en (Hugging Face-hosted raw clinical medicine-related textbook content) using MLX and again downstream using SNOMED-CT DRIFT (focusing on a relevant subset of its concept hierarchies) and smaller, proportional replays of common public training datasets for capabilities relevant for conversational medical expert systems (OpenHermes, billsum, etc.). The groups mentioned in the prompt are from the Centers for Medicare & Medicaid Services' Home Health Patient-Driven Groupings (PDGM) Model, which describes a Prospective Payment System used as an industry standard for reimbursing Medicare home care services that is very heavily ontological and statistical in its own right. It uses ICD to describe medical problems so these can be mapped to semantically equivalent SNOMED-CT concepts. The coloring scheme for the generated tokens uses MLX's utility for using colors to indicate each token's probability: white for more than 95%, green for less than that but more than 70%, and red for anything less than 30% The prompt parameters are: 2500 maximum tokens, 60% temperature, and 70% top-p The pdgm_group_with_comorbidity_definitions.txt file (provided as context to the model prompt) is generated from a script that, among other things, summarizes the semantics of the SNOMED-CT concepts in those groups by their definitional characteristics, which provides good In-context learning opportunity (i.e., RAG) for a model already trained to characterize medical concepts in a definitional manner. ---- % python -m mlx_tuning_fork.training -nt 2500 -t .6 -tp .7 --loom-file=sct_prompt.toml --build-prompt "system_prompt_dx_summary [pdgm_group_with_comorbidity_definitions.txt] cl_question" -f llama3 --adapter Llama-3-MrGrammaticaOntology-SCT-DRIFT-clinical-problems-0.5.0.npz/adapters.safetensors --loom-markers "question=In 4-6 paragraphs, summarize the Endocrine_4 and Endocrine_5 comorbidity groups in the provided context and describe what their findings and disorders have in common and how they are different according to SNOMED-CT" snomed_dist_train.yaml [..] Generating ========== Prompt: You are a helpful medical assistant who summarizes medical problems directly for a medical professional. Follow the instructions and by thinking step-by-step. Don't make up answers if you don't know. The context below are comorbidity groups of clinical diagnoses. Each includes an example in each group from SNOMED-CT. [..]

9