From Pixels to Images: How DICOM Stores Medical Images

Understanding Pixel Data, bit depth, and why medical images are more than just pictures.

Photo by Umanoide on Unsplash

Open any DICOM file, and you’ll find hundreds of pieces of information.

Patient name.

Study date.

Scanner manufacturer.

Image dimensions.

And eventually, you’ll come across one element with a very simple name:

Pixel Data.

This is where the actual image lives.

But here’s something that surprised me when I first started learning DICOM:

A medical image isn’t stored like a photo on your phone.

Instead, it’s stored as millions of numbers.

Every Image Is Just Numbers

When we look at a CT scan, we see shades of gray.

The computer doesn’t.

It only sees numbers.

Imagine a tiny image made of just four pixels.

12   45
180 255

Each number represents the brightness of a single pixel.

The viewer reads these values and turns them into the image we recognize.

Scale that up to a modern CT scan with hundreds of slices, and you begin to understand how much data is involved.

What Is Pixel Data?

In DICOM, the image itself is stored in a Data Element called Pixel Data.

Everything before it describes the image.

Pixel Data is the image.

Without it, you’d still know the patient’s name, the study date, and the scanner model — but there would be nothing to display.

More Than Black and White

Most everyday images use 8 bits per pixel.

That means each pixel can have 256 different values.

Medical imaging often needs much more precision.

A CT scan commonly uses 16 bits per pixel, allowing over 65,000 possible values.

Why?

Because tiny differences in tissue density can matter.

Having more possible values helps preserve those subtle differences.

Different Modalities, Different Pixel Values

Not every imaging modality stores pixel values in the same way.

A CT scan measures X-ray attenuation, producing values related to tissue density.

An MRI captures signals generated by magnetic fields and radio waves.

Ultrasound records reflected sound waves.

Even though they all use DICOM, the meaning of a pixel depends on the imaging modality.

That’s one reason DICOM stores so much metadata alongside the image.

Why Metadata Matters

Imagine receiving a matrix of numbers like this:

1024 1008 998 ...
1019 1003 995 ...
...

Without context, those values don’t tell you much.

Are they from a CT scan?

An MRI?

An X-ray?

How many rows and columns does the image have?

How many bits were used to store each pixel?

DICOM answers these questions with metadata.

The image and its description travel together.

Viewing Pixel Data with Python

Using pydicom, accessing the image is surprisingly simple.

import pydicom

ds = pydicom.dcmread("CT_Chest.dcm")

pixels = ds.pixel_array

print(pixels.shape)

The pixel_array is a matrix of numbers representing the image.

Libraries like NumPy make it easy to process these values for visualization, analysis, or AI applications.

Why This Matters for AI

If you’re working with machine learning, you’ll rarely train a model on the DICOM file itself.

Instead, your model learns from the pixel values.

Understanding where those values come from — and how DICOM stores them — is the first step toward building reliable medical imaging applications.

Key Takeaways

• The image itself is stored in the Pixel Data element.
• Medical images are collections of numerical pixel values.
• DICOM stores both the image and the metadata needed to interpret it.
• Different imaging modalities give different meanings to pixel values.
• Libraries like pydicom can convert Pixel Data into arrays that developers can work with.

Final Thoughts

When we open a CT scan, we naturally see an image.

Computers don’t.

They see a matrix of numbers.

One of DICOM’s greatest strengths is that it doesn’t just store those numbers — it also stores everything needed to understand what they represent.

That’s what transforms raw pixel values into meaningful medical images.

Before you go

• Please take a moment to like the post and follow the writer!
• Did you know that over 400,000 developers share what they’re building, learning, and discovering across our platforms every month? Learn how you can contribute here