Posts

Validating a Foundation Model Pipeline for Breast Cancer Histopathology: UNI, UMAP, and Spatial QC

Image
A whole-slide image (WSI) doesn't look like much to a neural network. It's a gigapixel file, often 100,000 by 100,000 pixels and tens of gigabytes on disk, and no architecture can ingest that as a single tensor. So the field works around it: slice the slide into thousands of small tiles, encode each tile into a compact feature vector, and treat the whole slide as a bag of those vectors. This post walks through the encoding step (UNI) and the validation checks that come right after it, in the context of a real pipeline predicting breast cancer biomarkers (ER, PR, HER2, Ki-67) from H&E-stained WSIs. Patch-based preprocessing: why tiling is unavoidable Tiling is the standard workaround for gigapixel inputs in digital pathology. In this pipeline, each slide is segmented for tissue (via HistoQC), then patched into 256×256 px tiles at 20x magnification using CLAM's preprocessing utilities. The output isn't raw image crops, it's coordinate metadata: an .h5 file per sl...

Artifical Intelligence in Breast Cancer Pathology

Image
  Introduction Breast cancer is the most commonly diagnosed cancer in women, and the biopsy slide is still the final word on it. That last step is also the slowest and the most subjective. Reading histopathology takes years of training, and even experienced pathologists disagree on the harder calls, especially at the fuzzy borders between benign, atypical, and malignant tissue. Caseloads keep rising while the number of specialists does not. That pressure is what has moved artificial intelligence from a research idea into a tool that increasingly sits beside the pathologist at the microscope. Looking across the recent literature, a few clear shifts stand out. From glass slides to digital ones                                       Figure: WSI image of breast cancer with 10x, 20x ,and 40x magnification. None of this happens without digitization. Whole slide imaging, which turns a glass sli...

Influenza Virus Evolution: Challenges of Antigenic Drift and Shift in Vaccine Design and Response

Image
  I ntroduction The under lying principles of influenza vaccination have not altered significantly over 70 years. The greatest challenge lies in the virus continuously evolving by antigenic drift and shift, reducing the efficacy of the vaccine and requiring ongoing revision. Curren tly, available vaccines rely on the tactic of attacking specific viral proteins, which are too few to defend against all the strains. Developing effective vaccines for influenza is complicated due to the virus’s propensity to mutate frequently, especially in the hemagglutinin (HA) protein. Even though antibodies to the HA head are known to neutralize infection, this region is highly variable, and therefore, vaccines are not very effective in the long term. Thus, vaccines currently in use offer temporary protection and must be updated annually. Influenza A and B viruses continue to cause severe disease across the globe, and evolutio n is tracked by the WHO for vaccine composition. Although this system en...

How Illumina Sequencing Works

Image
Illumina sequencing is the most widely used DNA sequencing platform today. It powers everything from cancer genomics to microbiome research, including antimicrobial resistance (AMR) surveillance. Overview of the Workflow The entire Illumina sequencing process is divided into four major stages:   Stage Name Purpose 1 Library Preparation Fragment DNA and attach adapter sequences to each fragment end 2 Cluster Generation Amplify each fragment into ~1,000 identical copies on the flow cell 3 Sequencing by Synthesis Add fluorescent nucleotides one at a time and image each cycle 4 Data Analysis Base-call reads and align to a reference genome or assemble de novo   Stage 1 : Li brary Preparation We start with a biological sample like tissue, blood, a swab, or environmental material, and extract genomic DNA. This raw DNA is far too long to sequence directly, so the first step is fragmentation: breaking it into short pieces, typically 150–300 base pairs long. Fragmentation is performed ...