Sunday, July 19, 2009

INTRODUCTION TO PATTERN ANALYSIS MODIFIED BY KITTLER

Use the Labels opposite to use the Algorithm

Several algorithms and systems have been devised to diagnose skin lesions, particularly pigmented skin lesions.
As these systems have evolved terminology has been accumulated to describe structures seen through the dermatoscope. The terminology has in the past often been metaphorical.
There are terms like ovoid nests, lagunes, lacunes, globules, leaf-like structures,maple leaf like structures, spoke-wheels, radial streaming, blue-white veil, peppering, fat fingures, sulci-gyri, crypts, milia-like cysts, arborosing vessels, etc etc. Often there are no precise definitions of these terms. Also, metaphorical terms have emotional connotations. That means that the meaning evoked for one person may be different to that evoked for another.


A Metaphoricoma



The Pattern Analysis modified by Kittler has simple defined terms for basic structures.
All structures can be described using the following 5 terms
Lines (Reticular, branched, curved, radial, parallel)
Dots
Clods
Circles
Pseudopods

Areas with none of the above are termed "structureless"

All patterns can be described as combinations of these structures
ALL OF THE STRUCTURES ARE ILLUSTRATED BELOW. THERE ARE NO OTHERS TO REMEMBER!










There are 2 parts to the Kittler algorithm
1. To Diagnose a Melanoma
2. Algorithm for Specific Diagnosis

The "To Diagnose a Melanoma" algorithm is arguably the simplest and most effective algorithm devised
If there is one pattern and one colour you move to the next lesion because that lesion is benign.
If there is more than one pattern OR more than one colour then you look for one of 8 "clues to melanoma". The discovery of just one clue may mandate the need for appropriate biopsy.
The clues are as follows
1. Eccentric structureless zone (any colour except skin colour)
2. Grey dots, clods, circles or lines
3. Black dots or clods, peripheral
4. Lines radial or pseudopods, segmental
5. Serpentine, helical and/or coiled vessels
6. White lines
7. Thick lines reticular or branched
8. Lines parallel, ridges

What about exceptions to the rule? What about nodular or amelanotic melanomas with one pattern and one colour?
Harald has stated categorically that his algorithm is not meant to be robotic. Everybody is encouraged to make his own modifications and develop his own algorithm.

I apply two principles whenever assessing skin lesions which override all other algorithmic decisions
1. If the lesion is EFG positive I biopsy it ( Elevated, Firm, Growing)
2. The "AORAKI" principal - If I can't make a CONFIDENT SPECIFIC BENIGN diagnosis I biopsy it (this is where the 2nd part of the Kittler algorithm is important. It is a tool for making a SPECIFIC benign diagnosis)

These two principles have yielded me amoung other things a Merkel cell carcinoma, an amelanotic melanoma, a nodular melanoma and several SCC,s that looked like seborrhoeic keratoses.

The second part of the Kittler algorithm "The Algorithm for Specific Diagnosis" is an elaborate systen for stepwise evaluation of lesions to identify them specifically ( not just as benign or suspicious). That is all condensed onto one poster available as a PDF file on this blog. You will need to have it printed at at least A3 size to read it as it represents 62 web pages condensed to one page.
This algorithm is not meant to be learned in one day! It is "complex but simple". It "falls together" when applied logically. Put the poster where you can refer to it often.

The Kittler method of describing vessels is based on the same simple principles with defined precise terminology that can be used to describe all vessels encountered.
Vessels can be
dots
clods
Lines straight
Lines looped
Lines curved
Lines serpentine
Lines helical
Lines coiled

and they can have patterns :-
Random pattern
Clustered
Serpiginous
Radial
Reticular
Branched



DEFINITIONS
Dots are tiny round spots not extending in any direction (no length and no breath)
Clods are well circumscribed, solid objects larger than dots that may take any shape (length and breath)
Lines are 2-dimensional continuous geometric figures extending in one direction (only length but no breath)
A Circle is a curved line equidistant from a central point
A Pseudopod is a line with a bulbous end


With Respect to Vessels
Linear vessels without a bend are straight.
Linear vessels with only one bend are looped if the bend is sharp (180°) and results in a reversal of direction (U-turn)
Linear vessels are curved if the bend is gentle.
Linear vessels with more than one bend are termed serpentine if the bends are arranged in a snakelike fashion
Linear vessels are Helical if the bends are twisted along a central axis
Linear vessels are Coiled if the bends are convoluted compactly (Figure 1).

Harald's own definitions are as follows.
straight= no bend
looped= one bend, 180 degrees
curved= one bend, less than 180 degrees
Serpentine= much more than one bend, any degree, not convoluted compactly
Coiled=more than one bend, convoluted compactly
Spiral= more than one bend, not convoluted compactly but convoluted along a central axis


The additional attributes thin and thick may be applied to all types of linear vessels. The attribute thick should only be applied if the thickness of the vessels exceeds the diameter of normal nail fold capillaries by far.
If one type of vessel is dominating by far the vascular pattern is termed monomorphic. If there are more than one type of vessel the pattern is termed polymorphic.
As a rule, a pattern is formed by multiple vessels that appear to be distributed in a random fashion and are not arranged in a specific way. However, it is important to note that there are some exceptions
If vessels that appear as dots or coiled vessels are not evenly distributed but concentrated in certain areas we call this pattern clustered.
If vessels that appear as dots or coiled vessels are arranged in a linear, arciform pattern we call that specific arrangement serpiginous ( eg Clear Cell Acanthoma)
Linear vessels (of the type straight, curved or looped) at the periphery that are orientated towards the centre of the lesion but are not crossing it are termed radial ( eg sebaceous hyperplasia and keratoacanthoma)
Linear straight vessels may cross each other and may be arranged in a way similar to pigmented lines reticular. We therefore call this vascular pattern reticular ( eg variant of mastocytosis known as telangiectasia macularis perstans)
Multiple serpentine vessels may divide from a main stem (usually a thick serpentine vessel), a pattern which is termed branched.



FINALLY - If Alan can do it it can't be that difficult...



"Sie haben eine Methode vor dem blinkenden!"


See the Labels opposite top to access the Algorithm

I have condensed Harald's entire algorithm ( all 62 Web pages) onto a single poster. You can go to http://www.skincancerconsult.com/ to the Presentations section under Word and PDF files -you will need to obtain a username and password from Ian McColl (imccoll@ozemail.com.au)- and download the PDF of the poster. Alternatively email me at cliffr@qld.chariot.net.au and I will happily send you the PDF file. It needs to be printed out at least on A3 paper, preferably glossy photo paper.
This algorithm is totally up to date. After Harald emailed me the changes he sent the following message:-
"The algorithm looks great. Thank you again for your work. I personally will not make any additional changes. Now its up to everybody out there to test it, change it, tear it apart, put it together again, and to add, remove,and change diagnoses."

3 comments:

Unknown said...

Has Harald cganged anything from the laminated poster that I bought at the Conference....Paul Buckley

Unknown said...

by the way, thanks for succinct overview of Kittler method...paul

Dr Cliff Rosendahl said...

Yes Paul, Harald made a few minor changes particularly to include pigmented IEC. The new version is available (free) as a downloadable PDF via a link to this blog or alternatively if you email me on cliffr@qld.chariot.net.au I will email you the PDF file. You can get it printed out from that file but as you know it needs to be at least A3 size or you can't read it!