---
title: Simetria Article Helpers
tags: Simetria, Article, Background
description: Notes
---

# Background Simetria!

# TO HELP ON INTRODUCTION
## Intracranial Hemorrhage (ICH)

ICH is a type of bleeding that occurs inside the skull (cranium). Bleeding caused by a blood vessel in the brain that has leaked or ruptured (torn) is called hemorrhagic stroke. (A bleeding within the skull is still referred to as ICH).

A hemorrhage generally happen suddenly, from either external or internal causes and can rapidly cause brain damage and it is usually life-threatening.

When a hemorrhage or stroke interrupts blood flow around or inside the brain, depriving it of oxygen for more than three or four minutes, the brain cells die and the affected nerve cells and the related functions they control are damaged as well.

- Intra-axial hemorrhage: Intracerebral hemorrhages may disrupt or compress adjacent brain tissue, leading to neurological dysfunction. Intra-axial hemorrhage are more dangerous and harder to treat than extra-axial bleeds. There are two major subtypes of intracerebral hemorrhages:

**Intraparenchymal** is blood that is located completely within the brain itself; Intraparenchymal hemorrhage accounts for approx. 8-13% of all strokes and results from a wide spectrum of disorders. It is more likely to result in death or major disability than ischemic stroke or subarachnoid hemorrhage, and therefore constitutes an immediate medical emergency.

**Intraventricular or subarachnoid** is blood that has leaked into the spaces of the brain that normally contain cerebrospinal fluid (the ventricles or subarachnoid cisterns).

- Extra-axial hemorrhage: Bleedings that occurs within the skull but outside of the brain tissue, falls into three subtypes

**Epidural** occurs when blood accumulates between your skull and the outermost covering of the brain. It typically follows a head injury, and usually with a skull fracture. High-pressure bleeding is a prominent feature. If you have an epidural hematoma, you may briefly lose consciousness and then regain consciousness.

**Subdural** is a collection of blood on the surface of the brain.

It’s typically the result of your head moving rapidly forward and stopping, such as in a car accident. However, it could also suggest abuse in children. This is the same type of movement a child experiences when being shaken.

**Subarachnoid** blood is seen layering into the brain along sulci and fissures, or filling subarachnoid cisterns (most often the chiasmatic cistern because of the presence of the anterior cerebral arteries of the circle of Willis and their branchpoints within that space). The classic presentation of subarachnoid hemorrhage is the sudden onset of a severe headache (a thunderclap headache). This can be a very dangerous entity, and requires emergent neurosurgical evaluation, and sometimes urgent intervention.

A subdural hematoma is more common than other ICHs in older people and people with history of heavy alcohol use.

<img src=https://d3i71xaburhd42.cloudfront.net/c022f4ba2938cdb2ef114b8d7a49e1670b6338c2/2-Figure1-1.png height="350" width="800">

<div style="text-align:center"><img src=https://www.radiologycafe.com/images/basics/[e-mail] height="500" width="500"></div>

While all acute (i.e. new) hemorrhages appear dense (i.e. white) on computed tomography (CT), the primary imaging features that help Radiologists determine the subtype of hemorrhage are the location, shape and proximity to other structures.

### References
[http://casemed.case.edu/clerkships/neurology/Web%20Neurorad/CT%20Basics.htm] - Basics of CT exams
[http://casemed.case.edu/clerkships/neurology/Web%20Neurorad/NR_Classic.html] - Bunch of sample for bleeding exams

[https://www.healthline.com/health/extradural-hemorrhage] - explain type and subtypes
[https://en.wikipedia.org/wiki/Intracranial_hemorrhage] - explain type and subtypes

[https://my.clevelandclinic.org/health/diseases/14480-intracranial-hemorrhage-cerebral-hemorrhage-and-hemorrhagic-stroke] - more about intracranial (with no differentiation in types and subtypes)
[https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/overview/hemorrhage-types] - kaggle walkthrough also shows a lot of explanations
[https://sci-hub.se/10.1016/S1474-4422(08)70041-3] - Imaging of intracranial haemorrhage

## DICOMS
DICOM® — Digital Imaging and Communications in Medicine — is the international standard to transmit, store, retrieve, print, process, and display medical imaging information. It defines the formats for medical images that can be exchanged with the data and quality necessary for clinical use.

DICOM groups information into data sets. Meaning each file contains patient information so that the image can never be separated from this information by mistake. A DICOM data object consists of a number of attributes, including items such as name, ID, etc., and also one special attribute containing the image pixel data. A single DICOM object can have only one attribute containing pixel data. For many modalities, this corresponds to a single image.

DICOM® is implemented in almost every radiology, cardiology imaging, and radiotherapy device (X-ray, CT, MRI, ultrasound, etc.), and increasingly in devices in other medical domains such as ophthalmology and dentistry. With hundreds of thousands of medical imaging devices in use, DICOM® is one of the most widely deployed healthcare messaging standards in the world. There are literally billions of DICOM® images currently in use for clinical care.

Since its first publication in 1993, DICOM® has revolutionized the practice of radiology, allowing the replacement of X-ray film with a fully digital workflow. Much as the Internet has become the platform for new consumer information applications, DICOM® has enabled advanced medical imaging applications that have “changed the face of clinical medicine”. From the emergency department, to cardiac stress testing, to breast cancer detection, DICOM® is the standard that makes medical imaging work — for doctors and for patients.

DICOM® is recognized by the International Organization for Standardization as the ISO 12052 standard.

### Who needs DICOM?

Physicians get better access to images and reports DICOM is in place. This allows them to make a faster diagnosis, potentially from anywhere in the world.

Patients can obtain faster and more effective care when DICOM is used to send their information through the healthcare enterprise.

Hospitals, clinics, imaging centers and specialists can make conformance to specific features of DICOM part of their purchasing requirements to ensure these tools work together across vendors to produce, manage and distribute images.

Manufacturers of imaging equipment (CT, MR, Ultrasound, etc), imaging information systems (HIS, RIS, PACS), and peripheral equipment (workstations, 3D printers, CD importers, etc) conform to DICOM to ensures compatibility of their equipment at every medical imaging facility.

Payers benefit from this faster and more effective process through potentially lowered cost of care.

### Image display

To promote identical grayscale image display on different monitors and consistent hard-copy images from various printers, the DICOM committee developed a lookup table to display digitally assigned pixel values. To use the DICOM grayscale standard display function (GSDF) [7] images must be viewed (or printed) on devices that have this lookup curve or on devices that have been calibrated to the GSDF curve [8].

### References
[https://www.dicomstandard.org/]
[https://en.wikipedia.org/wiki/DICOM#cite_note-7]
[https://sci-hub.se/10.1007/BF03168130] [7] - D. V. Smith et al. 1995 "Evaluation of the Medical Diagnostic Imaging Support System Based on 2 Years of Clinical Experience"
[https://sci-hub.se/10.1117/12.60322] [8]

## HOUNSFIELD SCALE
HU scale is a quantitative scale for describing radiodensity. The radiodensity of distiled water at STP (Standard Pressure and Temperature) is defined as zero HU, while the air at STP is defined as -1000 HU. In a voxel [1] $\mu$, the corresponding HU value is given by:

\begin{align}
HU = 1000 \frac{\mu - \mu_{water}}{\mu_{water} - \mu_{air}},
\end{align}

It is the definition for CT scanners that are calibrated with reference to water.

HU scale can range from -1000 (air) to +~2000 HU for very dense bone.

A practical application of this is in evaluation of tumors, where, for example, an adrenal tumor with radiodensity < 10 HU is rather fatty in composition and almost certainly a benign adrenal adenoma

The tissue surrounding a bleed is often less dense than the rest of the brain because of edema, and therefore shows up darker on the CT scan.

### References
[https://en.wikipedia.org/wiki/Hounsfield_scale]
[https://en.wikipedia.org/wiki/Voxel] [1] - Voxel
[https://sci-hub.se/10.1007/978-3-540-39408-2] - Buzug TM. "Computed Tomography: From Photon Statistics to Modern Cone-Beam CT". Springer; 2008:475 - Book
[Heymsfield S. Human Body Composition. Human Kinetics; 2005] - Book
[https://sci-hub.se/10.5681/joddd.2014.019] - Tahmineh Razi; MahdiNiknami; Fakhri Alavi Ghazani. "Relationship between Hounsfield Unit in CT Scan and Gray Scale in CBCT"

# TO HELP ON BACKGROUND
## KEYPOINTS AND DESCRIPTORS

Keypoints and descriptors are vectors that define a featured point in an image or scenery. To define them, images are taken under different variations of illumination, angle and scale space (Witkin,1983). Keypoint detection is actually a subset of blob detection, which aims to find interesting regions or spatial areas in an image.

### Scale Space
The scale space of an image is a function $L(x,y,σ)$. We obtain $L(x,y,σ)$ by convolution of a variable-scale Gaussian $G(x,y,σ)$ with the input image
\begin{align}
I(x,y):L(x,y,σ) =G(x,y,σ)∗I(x,y),
\end{align}

where ∗ is the convolution operator in x and y.

The Gaussian $G(x,y,σ)$ is given by

\begin{align}
G(x,y,σ) = \frac{1}{2\pi\sigma^2}e^{\frac{-(x^2 + y^2)}{2\sigma^2}}
\end{align}

Scale space images are obtained for different octaves, this is done by progressively blurring out images using Gaussian kernel. Lowe said that 5 blur and 4 octaves are enough for SIFT algorithm

<img src=https://www.oreilly.com/library/view/computer-vision-with/9781788299763/assets/e366ea0e-ed20-4660-be17-99f832786790.jpg height="700" width="650">

### Difference of Gaussians

The difference of Gaussian is computed at two nearby scales separated by a constant factor $k$

\begin{align}
D(x, y, \sigma) & = (G(x, y, k\sigma) - G(x, y, \sigma)) * I(x, y) \\
& = L(x, y, k\sigma) - L(x, y, \sigma)
\end{align}

<img src=https://www.researchgate.net/profile/Hannes_Fassold/publication/275042787/figure/fig5/AS:668367602077711@1536362662993/a-3D-DoG-scale-space-b-Detection-of-local-extrema-Images-courtesy-of-Lowe-et-al-1.ppm height="450" width="600">
For each octave, the image is convolved with Gaussians to produce the set shown on the left, adjacent Gaussian images are subtracted to produce the DoG shown on the right.

### Laplacian of a Gaussian
- Blur the image using Gaussian functions.
- Calculate the second order derivatives or the Laplacian - locates edges and corners (pilar for detecting keypoints).

The Laplacian of a Gaussian is represented as $\nabla^2 G$ and the scale-invariant Laplacian of a Gaussian would be $\sigma^2 \nabla^2 G$.

\begin{align}
\sigma^2 \nabla^2 G & = \frac{\partial G}{\partial \sigma} \approx \frac{G(x, y, k\sigma) - G(x, y, \sigma)}{k\sigma - \sigma} \\
& \therefore G(x, y, k\sigma) - G(x, y, \sigma) \approx (k-1) \sigma^2 \nabla^2 G
\end{align}

The DoG is approximately equivalent to the LoG

### Locate max-min in DoG images
Compare the pixel with its 26 neighbors (9+8+9). The pixel is selected only if is the larger/smaller of them all.

### Accurate keypoint localization
After each keypoint candidate is found, next step is to perform a detailed fit for location, scale and ratio. Points are rejected if they have low contrat (sensitive to noise) or are poorly localized along an edge. Brown and Lowe 2002 had developed a method for fitting a 3D quadratic function to the local sample points to determine the interpolated location of the maximum. This approach uses the Taylor expansion of the scale-space function $D(x, y, \sigma), shifted, so that the origin is at the sample point.

\begin{align}
D(x) &= D + \frac{\partial D^T}{\partial x}x + \frac{1}{2}x^T \frac{\partial^2 D}{\partial x^2}x,
\end{align}

where $D$ and its derivatives are evaluated at the sample point, and $x = (x, y, \sigma)^T is the offset from this point.

The location of the extremum, $\hat{x}$, is determined by taking the derivative of this function with rrespect to $x$ and stting it to zero, giving:

\begin{align}
\hat{x} = -\frac{\partial^2 D^{-1}}{\partial x^2} \frac{\partial D}{\partial x}.
\end{align}

If the offset $\hat{x} > 0.5$ in any dimension

<span style="color:blue">TO CONTINUE...</span>

### References
[https://www.cs.ubc.ca/~lowe/papers/ijcv04.pdf] - SIFT article
[https://www.cs.auckland.ac.nz/~rklette/CCV-Dalian/pdfs/E02_Features.pdf>]
[https://www.cs.utah.edu/~srikumar/cv_spring2017_files/Keypoints&Descriptors.pdf]
[https://en.m.wikipedia.org/wiki/Blob_detection]