PhD Scientific Days 2024

Poster Session H - Theoretical and Translational Medicine 1.

Quest for a Clinically Relevant Medical Image Segmentation Metric: the Definition and Implementation of Medical Similarity Index.

Előadó neve

Dr. Fazekas, Szuzina

Neptun code

JZINDU

Előadó munkahelye

Semmelweis University Medical Imaging Centre

Előadó telefonszáma

+36300981870

Előadó e-mail címe

fazekas.szuzina@semmelweis.hu

Az előadás címe

Quest for a Clinically Relevant Medical Image Segmentation Metric: the Definition and Implementation of Medical Similarity Index.

Szerző(k) neve és munkahelye

Szuzina Fazekas¹, Maurovich Horvat Pál², Zsolt Vizi³

1: Semmelweis Egyetem Orvosi Képalkotó Klinika Radiológiai Tanszék
2: Semmelweis University
3: University of Szeged

Bemutatás módja

Poszter

Szekció

Poster Session H - Theoretical and Translational Medicine 1.

Language of the presentation

English

Preferred session

Theoretical and Translational Medicine

Összefoglaló szövege

Introduction: In the field of radiology, an emerging number of images are created every day. In the medical field, delineation of different tissues and organs has a crucial role in diagnostics and therapeutics. The gold standard is manual segmentation by an expert, but nowadays there are more and more machine learning-based automatic segmentation methods. Thus, there is an understandable need for the quantification of the accuracy of a current segmentation. There are different widely used area-based and distance-based metrics, which are used for the evaluation of the accuracy of segmentations. These metrics only incorporate geometrical properties and fail to adapt to different clinical applications.

Aims: Our aim was to define and implement a clinically relevant medical image segmentation metric which has the opportunity to adapt to different clinical applications.

Methods: We use a reference contour, which we consider the gold standard segmentation, and we quantify the agreement of a test contour to the reference contour. We define bidirectional local distance, and based on this distance, we pair the points of the test contour to points of the reference contour. After correcting with the distance between the test and reference center of mass, we calculate the euclidean distance between the paired points and we give a score to each of the test points. The overall medical similarity index is calculated as the average of the scores along all the test points.

Result: We created an image processing pipeline in Python programming language. The code is available at a public GitHub repository, and we provide a runnable Google Colaboratory notebook. The algorithm can handle multislice images with more than one mask in one slice. We provide a mask splitting algorithm, which can separate the concave masks. Based on the evaluation of 140 neural network proposed masks of medical images, we fine tuned our metric properties.

Conclusion: We implemented a new segmentation evaluation metric [REF] and we provided a pipeline, which can be easily used for automatic measurement of clinical relevance of medical image segmentation.

Funding: "The scientific work was reached with the sponsorship of Gedeon Richter Talentum Foundation in framework of Gedeon Richter Excellence PhD Scholarship of Gedeon Richter."

University

Semmelweis University

Supervisor

Zsolt Vizi

Publication of my abstract

I do not give consent to the publication of my abstract on the website of the congress.

Kind

Szabad

Status

elfogadva

Accepted presentation method

poszter

Előadás fájl jóváhagyás

nem rendelkezett róla

Előadó

7239

Start

15:25

End

15:28