DeepPsy AG

Radiologists aren’t being replaced by AI - they’re being freed by it. With radiologist shortages and scan volumes skyrocketing, something had to give. The solution? AI that doesn’t replace clinicians but supercharges them. Generative and predictive AI are stepping in as quiet copilots. They are reviewing millions of scans, flagging urgent cases, and summarizing findings before the human expert even steps in. ➤ Aidoc’s AI has analyzed millions of CTs & X-rays, triggering early alerts that saved nearly 70 million minutes of review time ➤ PathAI brings machine learning to pathology, helping labs catch subtle patterns in complex biopsies ➤ Lunit Cancer Screening’s AI boosts cancer detection from mammograms and chest X-rays with sharp-eyed precision ✅ These tools don’t diagnose alone. They prioritize, assist, and flag ✅ They’re making radiology teams faster, more focused, and better supported ✅ The goal isn’t automation. It is augmentation with real clinical impact Radiology shows how AI and medicine can co-evolve. Not by sidelining clinicians, but by clearing the noise so their expertise shines where it matters most. 🧠 At DeepPsy AG, we’re learning from radiology’s lead. We believe psychiatry needs its own version of this precision shift—one rooted in the brain’s signals, not just symptoms. 🧠 Our platform (https://deeppsy.io/) decodes EEG biomarkers to guide treatment choices in: • SSRIs & SNRIs • rTMS • Ketamine • ECT Could psychiatry become the next radiology? Time to rethink how we see the brain. #AI #EEG #biomarkers #psychiatry #radiology #prediction #health #precision

Is your EEG "brain test" actually measuring blinks and muscle twitches instead of your brain? A new study from Roche (Denis A. Engemann) addressed a quiet issue in the EEG + machine learning world: Many models perform well, 𝐛𝐮𝐭 𝐨𝐧𝐥𝐲 𝐛𝐞𝐜𝐚𝐮𝐬𝐞 𝐭𝐡𝐞𝐲’𝐫𝐞 𝐫𝐞𝐥𝐲𝐢𝐧𝐠 𝐨𝐧 𝐧𝐨𝐧-𝐛𝐫𝐚𝐢𝐧 𝐬𝐢𝐠𝐧𝐚𝐥𝐬, 𝐥𝐢𝐤𝐞 𝐞𝐲𝐞 𝐦𝐨𝐯𝐞𝐦𝐞𝐧𝐭𝐬, 𝐟𝐚𝐜𝐢𝐚𝐥 𝐭𝐞𝐧𝐬𝐢𝐨𝐧, 𝐨𝐫 𝐡𝐞𝐚𝐫𝐭𝐛𝐞𝐚𝐭 𝐚𝐫𝐭𝐞𝐟𝐚𝐜𝐭𝐬. 𝑨𝒏𝒅 𝒘𝒉𝒆𝒏 𝒕𝒉𝒆𝒔𝒆 𝒂𝒓𝒕𝒆𝒇𝒂𝒄𝒕𝒔 𝒂𝒓𝒆 𝒓𝒆𝒎𝒐𝒗𝒆𝒅? 𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 𝒅𝒓𝒐𝒑𝒔. Meaning: what we thought was a brain biomarker might be something else entirely. ✅ 𝐁𝐨𝐝𝐲 𝐬𝐢𝐠𝐧𝐚𝐥𝐬 𝐜𝐚𝐧 𝐦𝐢𝐬𝐥𝐞𝐚𝐝 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧𝐬: EEG measures electricity from your head, but not all signals come from your brain. Eye blinks, jaw clenches, and even heartbeats can fool AI models, giving the illusion of accuracy. ✅ 𝐂𝐥𝐞𝐚𝐧 𝐄𝐄𝐆 𝐝𝐚𝐭𝐚 𝐦𝐚𝐤𝐞𝐬 𝐚 𝐛𝐢𝐠 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐜𝐞: The researchers found that when these non-brain signals were fully removed, many models became less accurate. This showed how much some models depended on body signals instead of brain waves. ✅ 𝐍𝐞𝐰 𝐦𝐞𝐭𝐡𝐨𝐝𝐬 𝐜𝐚𝐧 𝐟𝐢𝐱 𝐭𝐡𝐢𝐬: Using advanced techniques (Morlet wavelets and special mathematical models), the study showed we can make EEG biomarkers clearer and more accurate, based on true brain signals. 𝘍𝘰𝘳 𝘌𝘌𝘎 𝘵𝘰 𝘵𝘳𝘶𝘭𝘺 𝘵𝘳𝘢𝘯𝘴𝘧𝘰𝘳𝘮 𝘮𝘦𝘯𝘵𝘢𝘭 𝘩𝘦𝘢𝘭𝘵𝘩𝘤𝘢𝘳𝘦, 𝘸𝘦 𝘮𝘶𝘴𝘵 𝘣𝘦 𝘴𝘶𝘳𝘦 𝘸𝘦’𝘳𝘦 𝘮𝘦𝘢𝘴𝘶𝘳𝘪𝘯𝘨 𝘸𝘩𝘢𝘵 𝘳𝘦𝘢𝘭𝘭𝘺 𝘮𝘢𝘵𝘵𝘦𝘳𝘴: 𝘵𝘩𝘦 𝘣𝘳𝘢𝘪𝘯 𝘪𝘵𝘴𝘦𝘭𝘧. That’s why, at DeepPsy AG, we’re obsessed with making sure our biomarkers truly reflect brain function. Not noise. Not guesses. Actual biology. At DeepPsy AG, this #precision is our mission. 🧠 We build EEG tests using explainable AI to accurately identify which treatments (like antidepressants, magnetic stimulation, ketamine, or electroconvulsive therapy) will help specific patients, based on their actual brain signals. ✅ For clinicians, it means making better treatment choices: without guessing. ✅ For patients, it means quicker relief and fewer failed treatments. 👇 Want EEG biomarkers you can trust? Let’s connect. 👇 Let’s also 𝐜𝐨𝐧𝐧𝐞𝐜𝐭 𝐢𝐟 𝐲𝐨𝐮'𝐫𝐞 𝐢𝐧𝐯𝐞𝐬𝐭𝐢𝐧𝐠 𝐢𝐧 𝐭𝐡𝐞 𝐟𝐮𝐭𝐮𝐫𝐞 𝐨𝐟 𝐩𝐬𝐲𝐜𝐡𝐢𝐚𝐭𝐫𝐲, 𝐄𝐄𝐆, 𝐨𝐫 𝐧𝐞𝐮𝐫𝐨-𝐀𝐈. #EEG #Biomarkers #psychiatry #prediction #health #precision #AI

Flight attendant: Is there a doctor on board? Me (at DeepPsy AG ): Yes, but I’m not that kind of— Flight attendant: one patient in our plane hasn’t responded to 2 SSRIs, 1 round of rTMS, and is starting to lose hope. Their #EEG shows signal irregularities, and we need a prediction model.. Me: Okay. I’m here. 🧠 At DeepPsy AG , PhDs and MDs team up to bring #precision to psychiatry. ➤ MDs bring the clinical urgency ➤ PhDs bring the signal science Together, we’re decoding brainwaves to guide treatment—before trial and error begins. ✅ Not “that kind of doctor”? That’s exactly the kind we need. 🧠 Our platform analyzes EEG patterns to predict response to: • SSRIs & SNRIs • rTMS • Ketamine • ECT Because #psychiatry shouldn’t be guesswork: it should be brainwork. #psychiatry #EEG #prediction #health #biomarkers #precision

Can sleep #EEG predict who will respond to treatment in #depression? Most psychiatrists overlook this one nightly signal. But REM density - the frequency of rapid eye movements during REM sleep - may be one of our most promising biomarkers for depression. This landmark review by Steiger & Kimura shows how both wake and sleep EEG reflect brain states that predict treatment response and vulnerability to depression. ✅ 𝐊𝐞𝐲 𝐢𝐧𝐬𝐢𝐠𝐡𝐭𝐬: ➤ Shortened REM latency and increased REM density are robust markers across depression stages ➤ Persistent sleep EEG changes can remain after remission - a potential biological "scar" ➤ Elevated REM density is seen in high-risk individuals even before illness onset ➤ Prefrontal EEG cordance (a metric combining power and topography) correlates with antidepressant response ➤ Antidepressants often suppress REM sleep, but this is not always required for clinical improvement 𝐁𝐮𝐭 𝐭𝐡𝐞𝐫𝐞 𝐚𝐫𝐞 𝐫𝐞𝐚𝐥 𝐜𝐫𝐢𝐭𝐢𝐪𝐮𝐞𝐬 𝐭𝐨𝐨: ➤ Sleep EEG lacks specificity - similar patterns occur in other psychiatric disorders ➤ Most data are from lab settings, not real-world sleep environments ➤ EEG findings are influenced by age, gender, and medications, limiting generalizability 🧠 At DeepPsy AG, we are decoding the brain’s electrical signals to uncover who responds to which psychiatric treatments. 🧠 Our platform (𝘩𝘵𝘵𝘱𝘴://𝘥𝘦𝘦𝘱𝘱𝘴𝘺.𝘪𝘰/) analyzes electrophysiological signatures to identify predictive biomarkers of treatment response - across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT Follow DeepPsy AG for insights at the intersection of neuroscience, data, and compassionate care. We’re open to collaboration with researchers, EEG device makers, pharma partners, and all curious minds. If you're excited about meaningful innovation and scalable impact, let's connect. 📩 [email protected] #EEG #biomarkers #psychiatry #prediction #health #depression #precision

Could #EEG biomarkers actually change how psychiatrists treat #depression? After this conversation, we’re more convinced than before: yes. Yesterday, Mateo de Bardeci de Bardeci and I had the chance to visit Prof. Dr. Jochen Mutschler Mutschler at Luzerner Psychiatrie AG AG in Sarnen Together, we ➤ discussed how EEG (electroencephalography) could serve as a practical tool in everyday psychiatry ➤ explored how certain brainwave patterns might predict who responds to treatments like SSRIs, rTMS, or ketamine ➤ And most of all - we asked how this science could move from research papers into real clinical rooms Our vision was clear: This isn't just theory. With the right tools, this can become part of a psychiatrist's daily practice. And that changes everything. ✅ For patients, it means faster, more accurate treatment - fewer years lost in trial and error ✅ For psychiatrists, it offers data-driven support for decisions that often rely on guesswork Hopefully sooner than later, we will be witnessing a 𝐬𝐡𝐢𝐟𝐭 𝐟𝐫𝐨𝐦 𝐬𝐲𝐦𝐩𝐭𝐨𝐦-𝐛𝐚𝐬𝐞𝐝 𝐩𝐬𝐲𝐜𝐡𝐢𝐚𝐭𝐫𝐲 𝐭𝐨𝐰𝐚𝐫𝐝 𝐛𝐫𝐚𝐢𝐧-𝐛𝐚𝐬𝐞𝐝 𝐩𝐫𝐞𝐜𝐢𝐬𝐢𝐨𝐧 𝐜𝐚𝐫𝐞. These kinds of conversations - across disciplines, across perspectives - are exactly what move this field forward. Our platform (https://deeppsy.io/) analyzes electrophysiological signatures to identify predictive biomarkers of treatment response - across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT Follow DeepPsy AG for insights at the intersection of neuroscience, data, and compassionate care. #EEG #psychiatry #biomarkers #prediction #health #depression #precision

🚨 Precision psychiatry is here and DeepPsy AG is now open to investors. We are using EEG biomarkers + explainable AI to predict who responds to which treatment in depression. ✅ No more trial-and-error ✅ Faster relief, tailored care Join us: https://deeppsy.io/ #EEG #psychiatry #biomarkers #AI #investing #health

50 years on the same drug? ADHD deserves better. Despite ADHD being one of the most studied psychiatric conditions, our treatment approach has barely changed since the 1970s. Data from #EEG, fMRI, and genetics is shifting the narrative - hinting at a future where treatment is tailored, not trial-and-error. Here’s what’s breaking through the noise ✅ Promising Predictive Biomarkers for ADHD Treatment ➤ EEG leads the pack: Low-cost, accessible brainwave measures like individual alpha peak frequency (iAPF) and event-related potentials (e.g., P3 amplitude) consistently predict treatment response to stimulants, guanfacine, neurofeedback and more. ➤ Genetic signals are emerging: Polygenic risk scores (PRS) for ADHD explain ~2% of medication response; rare copy number variants (CNVs) within glutamate receptor genes predict better response to novel agents like fasoracetam. ➤ MRI is rich but impractical: Functional connectivity and right inferior frontal gyrus (IFG) activation show promise, but high cost, motion sensitivity, and lack of thresholds limit real-world utility. 🚨 But here’s the catch: ➤ Findings are rarely replicated. Most studies involve <100 participants, often White boys. ➤ Key brain-behavior links are correlational and often do not outperform basic clinical data. ➤ No clear biomarker thresholds exist for use in treatment allocation. We’re standing on the edge of precision psychiatry for ADHD. However, to cross over, we need bigger, more diverse samples, prospective validation, and true real-world implementation science. 🧠 At DeepPsy AG, we are decoding the brain’s electrical signals to uncover who responds to which psychiatric treatments. 🧠 We use cutting-edge software to bring precision to mental health care. 🧠 Our platform (https://deeppsy.io/) analyzes electrophysiological signatures to identify predictive biomarkers of treatment response - across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT 👇 Let’s connect if you're working on the future of psychiatry, EEG or neuro-AI. Follow DeepPsy AG for insights at the intersection of neuroscience, data, and compassionate care. #ADHD #EEG #Biomarkers #psychiatry #prediction #health #precision #depression #neuroimaging #genetics #mentalhealth

99% Accuracy Detecting Depression from #EEG? Here’s the Problem... A sweeping review of 92 studies on EEG-based depression diagnosis reveals an electrifying trend: AI is closing in on psychiatry’s most elusive challenge: objective biomarkers of mental illness. But here's the catch: beneath the 90%+ accuracies lies a maze of methodological gaps, unstandardized datasets, and untested generalizability. ➤ 𝐄𝐄𝐆 + 𝐀𝐈 = 𝐝𝐢𝐚𝐠𝐧𝐨𝐬𝐭𝐢𝐜 𝐩𝐨𝐭𝐞𝐧𝐭𝐢𝐚𝐥 Machine Learning (ML) models like SVM and KNN, and Deep Learning (DL) models like CNN-LSTM, consistently show 90–99% accuracy in classifying Major Depressive Disorder (MDD) from EEG signals. ➤ 𝐍𝐨𝐧-𝐥𝐢𝐧𝐞𝐚𝐫 𝐟𝐞𝐚𝐭𝐮𝐫𝐞𝐬 𝐥𝐞𝐚𝐝 𝐭𝐡𝐞 𝐰𝐚𝐲 Entropy, fractal dimensions, and phase synchrony outperformed simple frequency-based markers—hinting that depression scrambles not just activity levels, but the structure of brain dynamics. ➤ 𝐒𝐦𝐚𝐫𝐭𝐞𝐫 𝐰𝐢𝐭𝐡 𝐟𝐞𝐰𝐞𝐫 𝐜𝐡𝐚𝐧𝐧𝐞𝐥𝐬 High-performing models were often built from just 3–5 electrodes—suggesting future wearables could be both accurate and accessible. ➤ 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐢𝐬 𝐜𝐨𝐦𝐢𝐧𝐠 Reusing pre-trained models (e.g., from image recognition) could speed up EEG analysis and reduce the need for large datasets—still a major bottleneck. ⚠️ But the field isn’t ready for clinical deployment yet ➤ Overfitting is rampant due to small, homogeneous datasets ➤ Most models aren’t tested across multiple sites or populations ➤ Interpretability remains a weak link, especially for DL models 𝐓𝐡𝐢𝐬 𝐫𝐞𝐯𝐢𝐞𝐰 𝐞𝐱𝐩𝐨𝐬𝐞𝐬 𝐭𝐡𝐞 𝐟𝐚𝐮𝐥𝐭 𝐥𝐢𝐧𝐞𝐬 𝐛𝐞𝐭𝐰𝐞𝐞𝐧 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐩𝐫𝐨𝐠𝐫𝐞𝐬𝐬 𝐚𝐧𝐝 𝐜𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐭𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐢𝐨𝐧. 𝐈𝐭’𝐬 𝐚 𝐜𝐚𝐥𝐥 𝐟𝐨𝐫 𝐦𝐨𝐫𝐞 𝐫𝐨𝐛𝐮𝐬𝐭 𝐝𝐚𝐭𝐚, 𝐫𝐞𝐚𝐥-𝐰𝐨𝐫𝐥𝐝 𝐭𝐞𝐬𝐭𝐢𝐧𝐠, 𝐚𝐧𝐝 𝐦𝐨𝐝𝐞𝐥𝐬 𝐭𝐡𝐚𝐭 𝐜𝐥𝐢𝐧𝐢𝐜𝐢𝐚𝐧𝐬 𝐜𝐚𝐧 𝐭𝐫𝐮𝐬𝐭, 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐚𝐝𝐦𝐢𝐫𝐞. 🧠 At DeepPsy AG, we are decoding the brain’s electrical signals to uncover who responds to which psychiatric treatments. 🧠 We use cutting-edge software to bring precision to mental health care. 🧠 Our platform (https://deeppsy.io/) analyzes electrophysiological signatures to identify predictive biomarkers of treatment response—across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT Follow DeepPsy AG for more at the crossroads of EEG, AI, and psychiatry. #EEG #AI #Biomarkers #psychiatry #prediction #health #precision #depression

Can your brainwaves (EEG) reveal your #sexual orientation? A deep learning model says... maybe. A provocative #EEG study used deep learning to classify resting-state brain activity between homosexual and heterosexual men - with 83% accuracy. But here’s the catch: Pretrained networks that reliably classify sex from EEG failed to distinguish sexual orientation. Only a newly trained network could tease apart the patterns. ➤ Key findings: ✅ A deep neural net trained on sexual orientation (not sex) classified homosexual vs. heterosexual men from resting EEG with 83% accuracy. ✅ The most informative regions? Brodmann areas 1 and 40 - linked to somatosensory processing and attention. ✅ Peak frequencies: Not alpha, but beta (17 Hz) and gamma (48 Hz) activity, suggesting deeper network-level differences. ✅ Grad-CAM + eLORETA pinpointed spatiotemporal features driving classification - without any hand-crafted features. ✅ Homosexual men weren’t classified as “feminized” by the sex-trained model, challenging old assumptions. ➤ 3 points of critique: Small training sample (n = 77) with no external test set: limits generalizability. Resting-state EEG is sensitive to many confounds (fatigue, attention), raising concerns about interpretability. Ethical concerns: How these findings are used, or misused, matters more than what the model shows. 🧠 At DeepPsy AG, we are decoding the brain’s electrical signals to uncover who responds to which psychiatric treatments. 🧠 We use cutting-edge software to bring precision to mental health care. 🧠 Our platform (https://deeppsy.io/) analyzes electrophysiological signatures to identify predictive biomarkers of treatment response - across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT We're bringing explainable AI and neural biomarkers into clinical practice, to help the physicians not just treat, but tailor care based on the brain itself. What’s really going on here? Not a “gay brain” vs. “straight brain,” but perhaps subtle variations in attentional or sensory integration networks, which are rooted in both biology and experience. Deep learning opens a door to exploring complex identity traits in the brain, with power and responsibility in equal measure. 👇 Let’s connect if you're working on the future of psychiatry, EEG or neuro-AI. Follow DeepPsy AG for insights at the intersection of neuroscience, data, and compassionate care. #EEG #Biomarkers #AI #psychiatry #prediction #health #deep #neuroscience #orientation #precision #ethics

Can AI read EEGs better than humans for diagnosing epilepsy? Not yet. But we are getting closer. Despite decades of effort, the dream of automated diagnosis of epilepsy from routine EEG without relying on human interpretation of seizure spikes is still out of reach. A sweeping systematic 𝐫𝐞𝐯𝐢𝐞𝐰 𝐨𝐟 37 𝐬𝐭𝐮𝐝𝐢𝐞𝐬 shows that AI-based tools can reach up to 100% accuracy… But there’s a catch. Here’s what’s really going on: ➤ 𝐌𝐨𝐬𝐭 𝐬𝐭𝐮𝐝𝐢𝐞𝐬 𝐬𝐮𝐟𝐟𝐞𝐫 𝐟𝐫𝐨𝐦 𝐡𝐢𝐠𝐡 𝐫𝐢𝐬𝐤 𝐨𝐟 𝐛𝐢𝐚𝐬, especially in patient selection and data validation. Think: comparing seizure patients to healthy controls, not real-world clinical uncertainty. ➤ ⚠️ 𝐃𝐚𝐭𝐚 𝐥𝐞𝐚𝐤𝐚𝐠𝐞 𝐰𝐚𝐬 𝐫𝐚𝐦𝐩𝐚𝐧𝐭: training and testing on overlapping EEG data segments inflates accuracy. Only 22% of studies avoided this error. ➤ 𝐓𝐡𝐞 𝐚𝐯𝐞𝐫𝐚𝐠𝐞 𝐬𝐭𝐮𝐝𝐲 𝐢𝐧𝐜𝐥𝐮𝐝𝐞𝐝 𝐣𝐮𝐬𝐭 54 𝐩𝐞𝐨𝐩𝐥𝐞. Only 6 had more than 100. That’s nowhere near enough for reliable machine learning—especially deep learning. 𝐒𝐨 𝐰𝐡𝐚𝐭 𝐝𝐨𝐞𝐬 𝐭𝐡𝐢𝐬 𝐭𝐞𝐥𝐥 𝐮𝐬? Real-world epilepsy diagnosis is messy. Patients don’t walk in as textbook cases. AI must be tested in these gray zones, not on cherry-picked datasets. EEG is still a goldmine of latent biomarkers, but to strike gold, we need rigorous pipelines, standardized data, and reproducible code. #Deep #learning might scale better, but only with thousands of high-quality EEGs and methods to prevent overfitting and leakage. Criticisms of the current literature: ❌ Study design doesn’t mirror clinical settings ❌ Manual EEG segment selection introduces subjective artifacts ❌ Lack of external validation means findings might not generalize Instead of chasing flashy accuracy numbers, this review calls for clinical realism, transparency, and methodological rigor. And that’s a good thing. 🧠 At DeepPsy AG, we are decoding the brain’s electrical signals to uncover who responds to which psychiatric treatments. 🧠 We use cutting-edge software to bring precision to mental health care. 🧠 Our platform analyzes electrophysiological signatures to identify predictive biomarkers of treatment response - across interventions like: • SSRIs & SNRIs • rTMS • Ketamine • ECT We're bringing explainable AI and neural biomarkers into clinical practice, to help physicians not just treat, but tailor care based on the brain itself. Let’s connect if you're working on the future of psychiatry, EEG or neuro-AI. Follow DeepPsy AG for insights at the intersection of neuroscience, data, and compassionate care. #EEG #Biomarkers #prediction #psychiatry #health #precision #depression #epilepsy