Session 08 · No-Code & Agentic AI for Life Sciences

Biomedical
Image Analysis

CNN Intuition · Radiology Workflows · AI vs. Human Accuracy

Md. Jubayer Hossain · Founder & CEO
DeepBio Limited · DeepBio Academy

No Code & Agentic AI for Life Sciences · Session 08 · June 2026

How Computers See

CNN Intuition: Seeing Cells

Convolutional Neural Networks (CNNs) process images by scanning for features at different scales.

  • Edges/Lines: Low-level spatial features.
  • Shapes/Textures: Mid-level motifs (e.g., cell boundaries).
  • Objects/Patterns: High-level concepts (e.g., nuclei, infiltrates).

🧠 Translation Invariance

A CNN can recognize a cancer cell whether it is in the top-left or bottom-right of a slide. This is critical for biomedical robustness.

The Global Challenge

The Radiology Crisis

📉 Pneumonia Impact

Kills 2.5 Million people per year. Leading cause of death for children under 5 globally.

🏥 Personnel Shortage

In LMICs (Low-Middle Income Countries), there is often < 1 radiologist per 1,000,000 people.

"The bottleneck in global health isn't just medicine; it's diagnostic expertise at scale."
Implementation

AI Triage Workflows

How AI assists radiologists without replacing them.

1. Pre-Read Triage

AI scans the incoming queue and flags urgent cases (e.g., Pneumothorax) for the radiologist to read first.

2. Second Opinion

AI highlights suspicious regions (bounding boxes) that a tired radiologist might miss during a long shift.

3. Auto-Quantification

Measuring lesion volume or cardiothoracic ratio automatically, saving minutes per case.

Case Study

CheXNet: AI vs. Human

Stanford (2017) paper: A 121-layer CNN trained on ChestX-ray14 dataset.

  • Result: Outperformed the average radiologist on all 14 diseases in the dataset.
  • F1 Metric: Achieved higher accuracy for pneumonia detection than 4 Stanford radiologists.

📊 Key Finding

The model didn't just "see" the lungs; it learned subtle textures indicative of fluid and consolidation that are hard for the human eye to consistently quantify.

Source: Rajpurkar et al., "CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning"
Summary

Session 08 Key Takeaways

  • CNNs use spatial filters to process biomedical images.
  • AI addresses the global shortage of diagnostic expertise.
  • Triage workflows prioritize critical patients in the queue.
  • CheXNet demonstrated radiologist-level accuracy.
  • Human + AI collaboration is the clinical gold standard.
  • No-code vision models can be trained for specific niche tasks.

Next Session: Building Bio-AI Apps — Deployment and the final Project Showcase.

No Code & Agentic AI for Life Sciences

Thank You

Get in Touch

Md. Jubayer Hossain

bio.link/hossainlab

© 2026 Md. Jubayer Hossain · No Code & Agentic AI for Life Sciences — Session 08