Software Tools for Manual Segmentation of Tomography Images Supporting Radiologist’s Personal Context

Abstract

The article addresses the issue of developing of software tools for manual segmentation of tomography images supporting radiologist’s personal content. Through an expert survey we have identified the requirements for such software tools from the doctors’ point of view as end users, as well as to the nomenclature and functionality of tools that implement these requirements. In order to meet the identified requirements, we have developed a solution based on a client-server architecture with a cloud access point. The nomenclature of tools for marking tomographic images implemented in the solution, as well as the methodology for working with them, fully complies with the identified requirements. The following functions have been developed and implemented: calculation of the volume of the region of interest, as well as three options for the semiautomatic segmentation of the image based on threshold, extreme points and neuron networks. All functions have customizable parameters and (or) implementation options, which provides flexibility in solving specific markup problems. Experimental studies have shown that the constructed service meets all the requirements put forward by radiologists and corresponds to the global level in terms of accuracy and speed (performance) of segmentation.

Keywords

• tomography images
• ct
• mri
• manual segmentation
• automatic segmentation
• personal context
• neural netwoks
• segmentation tools