159: What If Your AI Tool Is Lying: Hidden Bias in Pathology Algorithms
Send us a textWhat if the AI tools we trust for cancer diagnosis are not always correct? This episode of DigiPath Digest takes on the uncomfortable but critical question: can AI “lie” to us—and how do we verify its performance before adopting it in clinical practice?Highlights:[00:02:00] Foundation models in action: Deployment of a fine-tuned pathology foundation model for EGFR biomarker detection in lung cancer—reducing the need for rapid molecular tests by 43%.[00:08:41] Bone marrow AI misclassifications: Why automated digital morphology still struggles with consistency across leukemia and lymphoma cases.[00:14:45] Lossy DICOM conversion: How file format changes can subtly—but significantly—affect AI model performance.[00:21:45] Federated tumor segmentation challenge: Coordinating 32 international institutions to benchmark healthcare AI fairly across diverse datasets.[00:27:47] AI in gynecologic cytology: Reviewing AI-driven Pap smear screening—promise, limitations, and why rigorous validation remains essential.[00:32:27] Takeaway: Trust but verify—AI tools must be validated before they can support or replace clinical decisions.Resources from this EpisodeNature Medicine – Fine-tuned pathology foundation model for lung cancer EGFR biomarker detection.Scientific Reports (Germany) – Study on how DICOM conversion impacts AI performance in digital pathology.Federated Tumor Segmentation Challenge – Benchmarking AI across 32 global institutions.Acta Cytologica – Review on AI in gynecologic cytology and Pap smear screening.Support the showBecome a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!
--------
27:49
--------
27:49
158: Multimodal Magic AI’s Role in Lung & Prostate Cancer Predictions
Send us a textWhat if AI could predict cancer outcomes better than traditional methods—and at a fraction of the cost? In this episode, I explore how multimodal AI is reshaping lung and prostate cancer predictions and why integration challenges still stand in the way.Episode Highlights with Timestamps:[00:02:57] Agentic AI in toxicologic pathology – what it is and how it could orchestrate workflows.[00:05:40] Grandium desktop scanners – making histology studies more accessible and efficient.[00:08:03] Clover framework – a cost-effective multimodal model combining vision + language for pathology.[00:13:40] NSCLC study (Beijing Chest Hospital) – AI predicts progression-free and overall survival with high accuracy.[00:17:58] Prostate cancer prognostic model (Cleveland Clinic & US partners) – validating AI-enabled Pathomic PRA test.[00:23:35] Thyroid neoplasm classification – challenges for AI in distinguishing overlapping histopathological features.[00:34:49] Real-world Belgium case study – AI integration into prostate biopsy workflow reduced IHC testing and turnaround time.[00:41:03] Lessons learned – adoption hurdles, system integration, and why change management is essential for successful digital transformation.Resources from this EpisodeWorld Tumor Registry – A global open-access repository for histopathology images: World Tumor RegistryBeijing Chest Hospital NSCLC AI Prognostic Study – Prognosis prediction using multimodal models.Cleveland Clinic Pathomic PRA Study – Independent validation of AI-enabled prostate cancer risk assessment.Grandium Scanners – Compact desktop scanners for histology slides: Grandium.aiSupport the showBecome a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!
--------
28:50
--------
28:50
158: AI in Pathology: From Pixels to Patients with Dr. Anil Parwani
Send us a textHow can pathology labs keep up with rising cancer diagnoses when the workforce is shrinking? Dr. Anil Parwani believes the answer lies in digital pathology powered by AI—and in this episode, he shares how his team at Ohio State University is making it work today.Amid increasing demands and workforce shortages, pathology is embracing digital transformation. The Ohio State University, for instance, has scanned over 4.2 million slides since 2016, leveraging digital pathology for standardization and objectivity. Current AI applications aid in biomarker quantification, rare event detection, and tumor classification, with future innovations expected in virtual staining, 3D pathology, and large language model integration. While integration challenges remain, these digital tools are poised to augment, rather than replace, human expertise, allowing each institution to navigate its unique "digital pathology chasm" with available market solutions.Episode Highlights with Timestamps:[00:02:15] From glass slides to digital workflows: why the shift was inevitable.[00:05:40] Whole slide imaging: achieving diagnostic quality equal to traditional microscopy.[00:12:22] AI in action: biomarker quantification, rare event detection, and tumor classification.[00:18:50] The next challenge—integrating AI seamlessly into LIS.[00:24:05] Virtual staining and 3D pathology: cutting costs and expanding insights.[00:32:10] Large language models: chatbots as diagnostic assistants and education tools.[00:39:00] Why human expertise remains irreplaceable in complex cases.[00:44:15] Global disparities: how to democratize digital pathology adoption.[00:50:30] The future: autonomous AI-assisted diagnostics and precision medicine.Resources from this Episode:Epredia Digital Pathology: https://www.epredia.com/products/digital-pathology Ohio State University Wexner Medical Center: pathology.osu.edu Support the showBecome a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!
--------
39:29
--------
39:29
157: How Academic Pathology Programs Can Prepare for AI | UPMC Podcast
Send us a text“AI in Pathology Isn’t Coming — It’s Already Here. Are You Ready?”From confusion to clarity — that’s what this episode is all about. I sat down with Drs. Liron Pantanowitz, Hooman Rashidi, and Matthew Hanna to dissect one of the most important and comprehensive AI-in-pathology resources ever created: the 7-part Modern Pathology series from UPMC’s Computational Pathology & AI Center of Excellence (CPAiCE). This isn’t just another opinion piece — it's your complete guide to understanding, implementing, and navigating AI in pathology with real-world insights and a global lens.Together, we discuss:Why pathologists and computer scientists are often lost in translationHow AI bias, regulation, and ethics are being addressed — globallyWhat it really takes to operationalize AI in patient care todayIf you’ve ever asked, “Where do I even start with AI in pathology?” — this is your answer.🔍 Highlights & Timestamps00:00 – The importance of earned trust in AI 01:00 – Education gaps in AI for both pathologists & developers 03:00 – Why CPAiCE was built & the three missions it serves 07:00 – The seven-part series: a blueprint for AI literacy 10:00 – Making AI education accessible without losing technical integrity 13:00 – How this series is being used for global teaching (including by me!) 17:00 – Generative AI in creating figures vs. human-authored content 21:00 – Eye-opening global AI regulations that pathologists MUST know 24:00 – Ethics, bias & strategies to mitigate real clinical risks 30:00 – What’s next: CPAiCE’s mission to reshape pathology education & practice 34:00 – A teaser: the first CPAiCE textbook is on the way!📚 Resources from This Episode📰 Read the full series (open access!): Modern Pathology 7-Part AI Series: https://www.modernpathology.org/article/S0893-3952(25)00001-8/fulltext👨⚕️ UPMC’s Computational Pathology & AI Center of Excellence (CPAiCE) 🌍 Creative Commons licensing means YOU can reuse, remix & teach from these resources — just cite the source.Support the showBecome a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!
--------
38:52
--------
38:52
156: Digital Pathology and AI in Cancer Grading, T-Cell Imaging & Biomarkers
Send us a textCan AI Grade Cancer Better Than Us? The Truth About T-Cell Imaging, Biomarkers & Digital Pathology DisruptionYou think Saturday mornings are for coffee? Try diving into bone marrow morphology, organ donor kidney biopsies, and AI-driven metastasis detection at sunrise. That’s how I do it—and you’re invited to join.Welcome to another data-packed episode of DigiPath Digest, where we explore the latest frontier in digital pathology and AI. This time, I reviewed some of the most exciting recent abstracts spanning cancer grading, T-cell quantification, and AI agents in oncology decision-making.These studies aren’t just fascinating—they’re redefining what’s possible in diagnostics, especially in under-resourced areas where digital pathology can create game-changing access and efficiency.🔬 Highlights with Timestamps[00:04:00] Detecting Metastases with Vision Transformers A team from Leeds Teaching Hospital developed a model for identifying lymph node and omental metastases in ovarian and peritoneal cancers with 99.8% AUROC and 100% balanced accuracy—this isn’t hype; it’s real AI pre-screening that could reduce diagnostic strain on pathologists.[00:08:00] DeepHeme: Bone Marrow Smears Meet AI UCSF and Memorial Sloan Kettering collaborated on DeepHeme, an ensemble deep learning model that classifies bone marrow aspirate cells with expert-level accuracy. With over 30K training images and strong external validation, it outperforms humans in both speed and detail.[00:16:00] Multimodal AI for Head & Neck Cancer This review showcases how integrating radiology, histopathology, and genomics with AI enhances personalized treatment and prognosis. Spoiler alert: Multimodal > unimodal.[00:24:00] Real-Time Kidney Biopsy Evaluation via AI Shoutout to our Digital Pathology Place sponsor, Techcyte, for their AI-powered tool improving accuracy and halving the time it takes to evaluate frozen kidney biopsies. This is the kind of innovation we need in organ transplantation.[00:32:00] GPT-4 as an Oncology Agent? Heidelberg researchers created an autonomous AI agent using GPT-4 plus vision models and OncoKB to handle oncology case decisions with 91% accuracy. This isn’t ChatGPT guessing—it’s a hybrid system citing guidelines and performing complex reasoning.🧠 Resources From This Episode📰 Multiple Instance Learning for Metastases Detection in Ovarian Cancer – Cancers journal🧬 DeepHeme: Generalizable Bone Marrow Cell Classifier – Science Translational Medicine📚 AI in Head and Neck Cancer: A Multimodal Review – Cancers journal🧪 AI-Assisted Review of Donor Kidney Pathology – Techcyte & Digital Pathology Place demo🤖 Autonomous AI Agent for Oncology Decisions – Heidelberg Group🎙️ Podcast on GPT-4 agents with Dr. Nina Kolker🧵 Earrings mentioned in the livestream? Find them in the DPP StoreI’d love to hear your feedback, your projects, and what digital pathology means to you. You can always reach out through comments, LinkedIn, or email.Support the showBecome a Digital Pathology Trailblazer get the "Digital Pathology 101" FREE E-book and join us!
Aleksandra Zuraw from Digital Pathology Place discusses digital pathology from the basic concepts to the newest developments, including image analysis and artificial intelligence. She reviews scientific literature and together with her guests discusses the current industry and research digital pathology trends.
Ouça Digital Pathology Podcast, Isso não é uma sessão de análise, com Vera Iaconelli e muitos outros podcasts de todo o mundo com o aplicativo o radio.net
Brasil