Jeremy Bikman’s Post

Just published "By 2030, 100% of Radiologists’ Current Job will be Done by AI" Ever since Geoffrey Hinton predicted that AI could replace radiologists, debates have sprung up arguing why this won’t happen. Many memes and caricatures have mocked Hinton’s prediction, yet a significant portion of radiology is already transforming today. The arguments often settle on ethics as the hardest barrier, citing AI model hallucinations and the complexity of ethical decision-making. However, this is largely an oversimplification and often serves as a convenient excuse to dismiss the inevitability of AI’s role. #AI #Radiology #ArtificialIntelligence

𝐈𝐬 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐫𝐞𝐩𝐥𝐚𝐜𝐢𝐧𝐠 𝐨𝐮𝐫 𝐫𝐚𝐝𝐢𝐨𝐥𝐨𝐠𝐲 𝐬𝐭𝐚𝐫𝐬? 𝑴𝒆𝒅𝑫𝒂𝒕𝒂 𝑴𝒐𝒏𝒅𝒂𝒚 0️⃣ 5️⃣ Giovanni Cacciamani MD, FEBU et al. (2023) pose this question in their paper. You can find out what they conclude at the end.   For many, the use of AI is still a black box, especially when it comes to the question of when it will work and be used. Arvid Moritz DENKweit now briefly explains the process of how such AI is created.   The use of AI is always understood as an iterative process, i.e. a process that is built up in several steps and is constantly being further developed and improved: 1)   Pure training phase - input from humans 2)   Validate results after initial training - verification by humans 3)   AI as co-pilot - independent suggestions from AI & humans only intervene in the event of errors   ✨ Attention: future 4)   AI decides independently - humans only evaluate randomly This is exactly the opinion of the authors of the paper.   They compared two groups with regard to the quality of mpMRIs for detecting prostate cancer: radiologists and radiologists using the AI model. The authors found that the combination of radiologists and AI system made the best assessments. Summarizing these results: „𝘛𝘩𝘦𝘳𝘦𝘧𝘰𝘳𝘦, 𝘵𝘩𝘦 𝘳𝘦𝘢𝘭 𝘨𝘢𝘮𝘦-𝘤𝘩𝘢𝘯𝘨𝘪𝘯𝘨 𝘮𝘰𝘷𝘦 𝘤𝘰𝘶𝘭𝘥 𝘣𝘦 𝘢 ‘‘𝘱𝘢𝘳𝘵𝘯𝘦𝘳𝘴𝘩𝘪𝘱’’ 𝘣𝘦𝘵𝘸𝘦𝘦𝘯 𝘈𝘐 𝘴𝘺𝘴𝘵𝘦𝘮𝘴 𝘢𝘯𝘥 𝘩𝘦𝘢𝘭𝘵𝘩 𝘤𝘢𝘳𝘦 𝘱𝘳𝘰𝘧𝘦𝘴𝘴𝘪𝘰𝘯𝘢𝘭𝘴 𝘳𝘢𝘵𝘩𝘦𝘳 𝘵𝘩𝘢𝘯 ‘‘𝘳𝘦𝘱𝘭𝘢𝘤𝘦𝘮𝘦𝘯𝘵’’ 𝘰𝘧 𝘵𝘩𝘦 𝘭𝘢𝘵𝘵𝘦𝘳.” AI can be used to standardize and generalize processes, which leads to greater efficiency and quality compared to purely human evaluation. The aim is to increase the hit rate of AI and reduce treatment time. 𝐒𝐨 𝐰𝐡𝐚𝐭 𝐢𝐬 𝐜𝐫𝐮𝐜𝐢𝐚𝐥 𝐟𝐨𝐫 𝐬𝐮𝐜𝐜𝐞𝐬𝐬𝐟𝐮𝐥 𝐀𝐈 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐚𝐧𝐝 𝐮𝐬𝐞 𝐢𝐧 𝐡𝐞𝐚𝐥𝐭𝐡𝐜𝐚𝐫𝐞 𝐚𝐬 𝐚 𝐜𝐨-𝐩𝐢𝐥𝐨𝐭? ✅ Errors made by AI must be detected and corrected by humans, otherwise they can grow exponentially ✅ More and more AI operators will be needed who also have expertise ✅ Critical questioning of results (just like with internet websites) remains crucial to explain the plausibility of AI Are you already using AI applications in healthcare, and if so, which ones and for what purpose?     PS: As you can see, we also had a lot of fun discussing the topic and I look forward to more conversations!

Artificial Intelligence (AI) is transforming radiology and its significance in healthcare. Lets discuss some startups and recent advancements in this field: AI in Radiology: Automated Image Analysis: AI algorithms analyze medical images, saving time for radiologists. They distinguish normal from abnormal findings with high accuracy. Early Detection and Diagnosis: AI spots subtle differences, aiding early disease detection. Functional Imaging: AI enhances precision and informs treatment decisions. Cancer Monitoring: AI helps monitor tumors and assess treatment effectiveness. Workflow Efficiency: AI streamlines radiology tasks, allowing radiologists to focus on critical aspects. Startups Leading the Way: Qure.ai: Offers deep learning solutions for diagnosis and treatment. Zebra Medical Vision: Provides real-time analysis of clinical imaging data. Aidoc: Flags acute anomalies in real-time, improving efficiency. Kheiron Medical Technologies: Focuses on early breast cancer detection. Nines: Applies AI to teleradiology. Recent Advancements: AI-assisted Ultrasound Scanning: Enhances US scanning accuracy. Digital Radiography: AI aids X-ray interpretation and other techniques. Market Consolidation and Workflow Optimization: Automation maximizes efficiency in patient care. These innovations promise better diagnostics, streamlined workflows, and improved patient outcomes. Check out some free tools from Standard Center for Artificial intelligence in Medicine and Imaging https://lnkd.in/gTigWtZp #artificialintelligence #healthcare #radiology #startups #deeplearning #future A recent podcast in ABC.

This perfectly illustrates the headwinds facing most AI companies - hospitals (and large clinics) can hire AI engineers to develop their own algorithms and train them on their own data sets. An algorithm isn't a business; what else do you bring to the table? #healthcare #healthcareai #medicalai #ai #generativeai #medicalimaging #radiology #pacs https://lnkd.in/dMYPpk2b

🌟 AI Applications in Ultrasound: Discover the Companies Leading the Way🌟 The integration of AI in ultrasound is transforming diagnostics and patient care, and several companies are at the forefront of this innovation. In this carousel post, I highlight key players in the industry, including our very own SIMUT Medical, leveraging AI to reshape the future of ultrasound. Swipe through to learn about these innovative companies and their contributions to the ultrasound landscape. Let's delve into the fascinating ways AI is pushing the boundaries of medical imaging, including SIMUT Medical's groundbreaking solutions! Don't forget to share your thoughts and experiences on AI in ultrasound in the comments below. Let's keep the conversation going! 🌟 کاربردهای هوش مصنوعی در سونوگرافی: شرکتهای پیشرو را بشناسید 🌟 ادغام هوش مصنوعی در سونوگرافی در حال تغییر روشهای تشخیصی و مراقبت از بیماران است و چندین شرکت در خط مقدم این نوآوری قرار دارند. در این پست، به معرفی شرکتهای کلیدی در این صنعت پرداختهام، از جمله شرکت SIMUT که با استفاده از هوش مصنوعی آینده سونوگرافی را دگرگون میکند. با ورق زدن تصاویر، با این شرکتهای نوآور و نقش آنها در دنیای سونوگرافی بیشتر آشنا شوید. بیایید با هم نگاهی بیندازیم به روشهای جذابی که هوش مصنوعی در حال گسترش مرزهای تصویربرداری پزشکی است فراموش نکنید نظرات و تجربیات خود را در مورد هوش مصنوعی در سونوگرافی در بخش نظرات با من به اشتراک بگذارید #AIinHealthcare #Ultrasound #MedicalImaging #Innovation #SIMUTMedical

"DeepTek.ai, a Pune-based AI startup, is revolutionizing radiology workflow with its FDA-approved Augmento X-Ray AI solution for chest X-rays. The tool uses deep learning algorithms to detect abnormalities, reducing radiologist workload by up to 30-50% and improving reporting quality. With plans for global expansion and strategic partnerships, DeepTek is making significant strides in the world of AI and healthcare. #AI #Healthcare #DeepLearning #Radiology"

I wanted to share our latest article on workload reduction with AI tools for chest xrays with you. https://lnkd.in/ekfeaZFX Louis Lind Plesner has done a wonderful job of carefully looking at the SmartChest dataset again in an effort to understand how many xrays could be autonomously and safely be reported on by AI. Results are off label, but interesting. Workload reduction in excess of 50% of cases is possible at 98% sensitivity. But radiologists (at least not in Denmark) don’t have 100% sensitivity. Their sensitivity is 87%. If we lower the sensitivity of the algorithm to this level than 82% can be automated. Then the AI would miss as many cases with pathology as the radiologists but more of them would be clinically significant. Twice as many. So AI needs to have a higher sensitivity to achieve a similar level of critical misses. Once again humans are better at reading chest xrays! Achieving large scale automation of unremarkable AI for chest Xrays seems to be a question of how many mistakes we are willing to accept. Thanks to the team and participating partners for their support. Louis Lind Plesner Michael Andersen Mikael Boesen Mathias Willadsen Brejnebøl Christian Krag Lene Collatz Laustrup Olav Wendelboe Nielsen Catherine Jones RAIT Denmark Herlev og Gentofte Hospital Bispebjerg Hospital #radiology

MAKING RADIOLOGY AI SYSTEMS MORE UNDERSTANDABLE AND TRANSPARENT ⬇ This paper explores the crucial topic of making AI systems, particularly in radiology, more understandable and transparent. 1️⃣ The central argument is that AI's "black box" nature, where the decision-making process is opaque, poses significant challenges in healthcare where explanations are vital for trust, error correction, and knowledge creation. 2️⃣ The paper draws an interesting parallel between understanding AI and understanding natural laws in physics, suggesting that even the most complex AI systems can be made understandable. 3️⃣ It emphasizes the necessity of explanations not just for ethical and trust reasons but as a means to identify and correct errors, thus improving AI systems over time. 4️⃣ A considerable section is dedicated to the field of Explainable AI (XAI), discussing tools and methodologies that aim to make AI decision-making processes more transparent and understandable. 5️⃣ Despite the potential of XAI, the paper also acknowledges the fallibility of these methods and stresses the ongoing need for critical evaluation and improvement. ✍🏻 Eric Marcus, Jonas Teuwen. Artificial intelligence and explanation: How, why, and when to explain black boxes. European Journal of Radiology, Volume 173, 111393. February 23, 2024. DOI: 10.1016/j.ejrad.2024.111393 💡 From my perspective, the methodologies discussed, alongside the comparison to scientific understanding in physics, illuminate a path forward in developing AI tools that are not just powerful but also understandable and ethically sound. This work aligns with the broader need for AI systems that clinicians, patients, and regulators can trust, thereby paving the way for more widespread acceptance and integration of AI in healthcare. 💥 Join our community of 30.000 curious minds! ✅ Subscribe to our newsletter for your front-row seat to the latest groundbreaking studies. Get started here: https://lnkd.in/eR7qichj.

Discuss the real value of #articificalintelligence in #radiology in a #ISIMJournal editorial by Alrujaib and Bakheet https://lnkd.in/eCJhKVnd

Discuss the real value of #articificalintelligence in #radiology in a #ISIMJournal editorial by Alrujaib and Bakheet https://lnkd.in/eCJhKVnd