AMR.ALFAYOUMY
← HOME / PROJECTS / COVID-19 CT AND X-RAY DETECTION
RESEARCH · 2021 · Medical imaging ML

COVID-19 CT and X-ray Detection.

AI/ML Researcher
Cairo, Egypt
Source: COVID CT/X-ray report

A supervised machine-learning study for COVID-19 detection from chest CT and X-ray scans using feature extraction, classical classifiers, and ensemble voting.

CT accuracy
98.2%
Voting classifier
X-ray accuracy
99.6%
Voting classifier
Images
3,698
2,471 CT + 1,227 X-ray
// 02 — CHALLENGE
WHY IT MATTERED

The project explored whether chest CT and X-ray images could support fast, low-cost COVID-19 screening when PCR access was constrained.

The challenge was to get strong diagnostic classification performance without relying on heavyweight GPU-driven deep learning pipelines.

// 03 — APPROACH
HOW I BUILT IT

I structured two supervised classification workflows: one for CT scans and one for chest X-rays, each trained and tested on public medical-imaging datasets.

The models used image feature extraction across raw pixel intensities, HSV color histograms, Haralick texture features, and threshold adjacency statistics, then compared k-NN, SVM variants, Random Forest, MLP, and hard-voting ensembles.

// 04 — HIGHLIGHTS
KEY TAKEAWAYS
  • Built separate CT and X-ray classifiers for COVID-19 versus non-COVID detection
  • Achieved 98.2% CT accuracy and 99.6% X-ray accuracy with hard-voting ensembles
  • Compared classical feature-extraction pipelines against multiple supervised classifiers
  • Included a lightweight web-app testing flow for novel CT and X-ray inputs
// 05 — OUTCOMES
RESULTS AND LESSONS
  • Showed that engineered image features plus classical ML could produce strong COVID-19 screening results with modest compute needs.
  • Validated the approach across two imaging modalities instead of a single scan type.
  • Positioned the work as a low-cost screening aid for settings where broad PCR testing was difficult.
// 06 — STACK
THE TOOLS
Models
k-NNLinear SVMRBF SVMRandom ForestVoting ClassifierMLP
Features
Raw pixelsHSV histogramsHaralick texturesThreshold adjacency statistics
Data
Chest CT scansChest X-raysOpenCVPython
// 07 — PROJECT PDF
ORIGINAL REPORT