https://www.scottpoulton.com/categories/histohelper/ Recent content in HistoHelper on somewhere/nowhere Hugo -- gohugo.io en-us Tue, 07 Apr 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/histohelper_case_study/ Tue, 07 Apr 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/histohelper_case_study/ HistoHelper: WSI Inference & Predictive Morphology 1. Tl;dr HistoHelper is an active research prototype exploring predictive morphology in oncology. The objective is to evaluate whether a Convolutional Neural Network can identify morphological signatures within primary skin lesions to predict specific organ tropism (e.g., pulmonary vs. cerebral metastasis). Frontend Interface: FastAPI, Uvicorn, Docker, Google Cloud Run Machine Learning Pipeline: PyTorch (MPS Hardware Accelerated), OpenCV, OpenSlide Model Architecture: ResNet18 (Feature Extraction) transitioning to Multiple Instance Learning. https://www.scottpoulton.com/posts/project-log-005-formalising-architecture/ Wed, 01 Apr 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/project-log-005-formalising-architecture/ Formalising the Architecture: ResNet and Predictive Morphology As HistoHelper shifts from a conceptual MVP into a structured research tool, it is necessary to formally ground the architecture in established literature. The primary objective remains unchanged: to determine whether a Convolutional Neural Network (CNN) can identify morphological signatures within primary melanoma Whole Slide Images (WSIs) that predict specific organ tropism (e.g., pulmonary versus cerebral metastasis). The Rationale for Residual Networks (ResNet) Training deep CNNs from scratch on gigapixel histological data is computationally prohibitive and highly prone to vanishing gradients. https://www.scottpoulton.com/posts/project-log-004-tensors/ Mon, 30 Mar 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/project-log-004-tensors/ Tensors, Transfer Learning, and Apple Silicon In Log 3, I mapped out the problem: processing 3GB .svs Whole Slide Images to predict melanoma organ tropism. The reality? Trying to build the complex tiling logic and the neural network architecture at the exact same time on a local machine is a recipe for a crashed Mac and zero progress. It was time for a strategic pivot. The MVP Pivot: SPIDER-Skin Instead of wrestling gigapixel monsters, I bypassed the tiling phase entirely to unblock the MVP. https://www.scottpoulton.com/posts/project-log-001-defining/ Wed, 25 Mar 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/project-log-001-defining/ Defining HistoHelper AI The Genesis Every good tool starts with a problem that is too annoying to ignore. In clinical pathology, the problem is volume and visual fatigue. Pathologists are tasked with scanning gigapixel SVS (Whole Slide Images) to find microscopic anomalies. It is finding a needle in a digital haystack, and visual fatigue is a real threat to diagnostic accuracy. The Problem Beyond basic anomaly detection, there is a massive, unanswered question in oncology known as Organ Tropism. https://www.scottpoulton.com/posts/project-log-002-histohelper/ Wed, 25 Mar 2026 00:00:00 +0000 https://www.scottpoulton.com/posts/project-log-002-histohelper/ The Needle in the Gigapixel Haystack I am officially starting work on HistoHelper. If you have read my About page, you know why I am in Medical Laboratory Science. I am here to understand the mechanics of disease, specifically melanoma, from the ground up. But understanding the theory is only half the battle. The other half is building tools that actually do something useful with that theory. The Problem with SVS Files In digital pathology, tissue biopsies are scanned into Whole Slide Images (WSIs), usually saved as .