diplomsko delo
Tajda Kreč (Author), Tina Skuhala (Author), Erna Huskić (Reviewer), Janez Žibert (Mentor), Rebeka Viltužnik (Co-mentor)

Abstract

Uvod: Umetna inteligenca je doživela velik napredek in se vse pogosteje uporablja v medicinski diagnostiki. Ima pomembno nalogo pri analizi medicinskih slik. Pri CT prsnih organov je ključnega pomena pravočasno odkrivanje pljučnih nodulov, saj lahko zgodnja diagnostika bistveno izboljša izid zdravljenja pljučnega raka. Vključevanje umetne inteligence v ta proces lahko prinese večjo natančnost pri prepoznavanju pljučnih nodulov, skrajša čas obdelave slik in zmanjša delovno obremenitev medicinskega osebja. Namen: Glavni cilj te diplomske naloge je raziskati možnosti uporabe umetne inteligence pri analizi CT slik prsnih organov, predvsem pri avtomatski detekciji pljučnih nodulov in segmentaciji pljuč. Osredotočamo se na odprtokodna orodja Monai Label in 3D Slicer, ki omogočata vzpostavitev in testiranje modelov umetne inteligence za analizo medicinskih slik. Poleg preučujemo tudi praktične izzive pri uporabi teh orodij ter razmišljamo o možnostih njihovega vpeljevanja v klinično prakso. Metode dela: Najprej smo opravili pregled obstoječe literature na področju umetne inteligence v radiologiji, pri čemer smo se osredotočali na tuje podatkovne baze. Nato pa smo izvedli študijo primera, kjer smo uporabili javno dostopne CT posnetke iz baze TCIA. Preizkusili smo model Lung nodule CT detection, ki je bil naučen na podatkovnem naboru LUNA16. S pomočjo programa 3D Slicer in razširitve Monai Label smo izvedli avtomatsko detekcijo pljučnih nodulov ter primerjali rezultate z javno dostopnimi ocenami radiologov. Z algoritmom Auto3DSeg smo izvedli segmentacijo pljuč na enakih primerih. Rezultati: V rezultatih smo ugotovili, da umetna inteligenca uspešno identificira pljučne nodule, vendar se v nekaterih primerih pojavljajo odstopanja v primerjavi z ocenami radiologov. To potrjuje, da so sistemi umetne inteligence lahko uporabni pri analizi CT slik, vendar jih je treba še dodatno izboljšati, preden jih vključimo v rutinsko klinično diagnostiko. Razprava in zaključek: Uporaba umetne inteligence pri analizi medicinskih slik odpira nove možnosti za hitrejšo in bolj natančno diagnostiko pljučnih bolezni. Kljub pozitivnim rezultatom, ki jih ponujajo odprtokodna orodja, kot sta Monai Label in 3D Slicer, so še vedno prisotni izzivi, kot so lažno pozitivni rezultati in potreba po dodatnem preverjanju s strani strokovnjakov. Vpeljevanje umetne inteligence v klinično prakso zahteva tudi ustrezno izobraževanje medicinskega osebja in prilagoditve obstoječih delovnih procesov.

Keywords

diplomska dela;radiološka tehnologija;umetna inteligenca;računalniška tomografija;pljučni noduli;segmentacija;avtomatska detekcija;3D Slicer;Monai Label;TCIA;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL ZF - University College of Health Studies
Publisher: [T. Kreč
UDC: 616-07
COBISS: 247471107 Link will open in a new window
Views: 87
Downloads: 13
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: ǂThe ǂuse of artificial intelligence in CT image processing of the Lung and detection of pulmonary nodules
Secondary abstract: Introduction: Artificial intelligence has made great progress and is increasingly being used in medical diagnostics. It plays an important role in the analysis of medical images. In chest CT, early detection of lung nodules is crucial, as timely diagnosis can significantly improve the outcome of lung cancer treatment. The integration of Artificial intelligence into this process can enhance accuracy in nodule identification, reduce image processing time, and lighten the workload for medical staff. Purpose: The main objective of this thesis is to explore the possibilities of using Artificial intelligence in the analysis of chest CT images, particularly in the automatic detection of lung nodules and lung segmentation. We focus on the open-source tools Monai Label and 3D Slicer, which enable the development and testing of Artificial intelligence models for medical image analysis. Additionally, we examine the practical challenges in using these tools and consider the possibilities of their implementation in clinical practice. Methods: First, we conducted a review of the existing literature on Artificial intelligence in radiology, focusing on foreign databases. We then carried out a case study using publicly available CT images from the TCIA database. We tested the Artificial intelligence model Lung Nodule CT Detection, which was trained on the LUNA16 dataset. Using the 3D Slicer program and the Monai Label extension, we performed automatic detection of lung nodules and compared the results with publicly available radiologist assessments. Additionally, we used the Auto3DSeg algorithm to perform lung segmentation on the same cases. Results: In the results, we found that Artificial intelligence successfully identifies lung nodules, but in some cases, discrepancies were observed when compared to radiologist assessments. This supports the idea that Artificial intelligence systems can be useful in CT image analysis, but further improvements are needed before they can be incorporated into routine clinical diagnostics. Discussion and conclusion: The use of Artificial intelligence in medical image analysis opens up new possibilities for faster and more accurate diagnosis of lung diseases. Despite the positive results offered by open-source tools such as Monai Label and 3D Slicer, challenges remain, including false positive results and the need for additional verification by experts. The integration of Artificial intelligence into clinical practice also requires proper training of medical staff and adjustments to existing workflows.
Secondary keywords: diploma theses;radiologic technology;artificial intelligence;Computed tomography;lung nodules;segmentation;automatic detection;3D Slicer;Monai Label;TCIA;
Type (COBISS): Bachelor thesis/paper
Study programme: 0
Thesis comment: Univ. v Ljubljani, Zdravstvena fak., Oddelek za radiološko tehnologijo
Pages: 1 spletni vir (1 datoteka PDF (33 str.))
ID: 27322206