Efficient and Precise Classification of CT Scannings of Renal Tumors Using Convolutional Neural Networks

Mikkel Pedersen, Henning Christiansen*, Nessn H. Asawi

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review


We propose a new schema for training and use of deep convolutional neural networks for classification of renal tumors as benign or malign from CT scanning images. A CT scanning of a part of the human body produces a stack of 2D images, each representing a slice at a certain depth, and thus comprising a 3D mapping. An additional temporal dimension may be added by injection of contrast fluid with CT scannings performed at certain time intervals. We reduce dimensionality – and thus computational complexity – by ignoring depth and temporal information, while maintaining an ultimate accuracy. Classification of a given scan is done by majority voting over the classifications of all its 2D images. Images are divided into training and validation sets on a patient basis in order to reduce overtraining. Current experiments with scans for 369 patients, yielding almost 20,000 2D images, demonstrate an accuracy of 93.3% for single images and 100% for patients.
Original languageEnglish
Title of host publicationFoundations of Intelligent Systems : 25th International Symposium, ISMIS 2020, Graz, Austria, September 23-25, 2020, Proceedings
EditorsDenis Helic, Martin Stettinger, Alexander Felfernig, Gerhard Leitner, Zbigniew W. Ras
Number of pages8
Publication date2020
ISBN (Print)978-3-030-59490-9
ISBN (Electronic)978-3-030-59491-6
Publication statusPublished - 2020
Event25th International Symposium on Methodologies for Intelligent Systems: Foundations of Intelligent Systems - Graz University of Technology - ONLINE, Graz, Austria
Duration: 20 May 202022 May 2020
Conference number: 25


Symposium25th International Symposium on Methodologies for Intelligent Systems
LocationGraz University of Technology - ONLINE
OtherISMIS (this year, it is organized as an online event) is an established and prestigious conference for exchanging the latest research results in building intelligent systems. It provides a basis for exchanging research results and transport scientific achievements towards industrial applications. The scope of ISMIS is to present a wide range of topics related to the application of Artificial Intelligence techniques related to areas such as decision support, knowledge representation, logical programming, knowledge-based systems, machine learning, planning, computer vision, information retrieval, configuration and diagnosis. The conference also focus on interdisciplinary research in AI-related fields, for example, decision support systems and human decision making or recommender systems and human personality, and knowledge-based systems development and cognitive aspects of knowledge understanding.Conference ScopeMotivated by recent developments in sub-symbolic AI and the continuous emergence of new application domains, this year’s conference theme is “Towards explainable Artificial Intelligence”. We this focus, ISMI2020 contributes to emerging challenges related to the explainability of system outputs which experience an increased relevance in areas such as autonomous driving, intelligent sales assistants, and different further application domains such as medicine, intelligent maintenance, and eLearning.ISMIS 2020 is intended to attract individuals who are actively engaged both in theoretical and practical aspects of intelligent systems. The goal is to provide a platform for a useful exchange between theoreticians and practitioners, and to foster the cross-fertilization of ideas: Relevant conference topics include but are not limited to:Explainable AI (XAI)Machine LearningDeep learningData MiningRecommender SystemsConstraint based systemsAutonomous systemsApplications (Configuration, Internet of Things, Financial Services, e-Health…)Intelligent user interfacesUser ModelingHuman computationSocially-aware systemsAutonomous systemsDigital librariesIntelligent AgentsInformation RetrievalNatural Language ProcessingKnowledge IntegrationKnowledge VisualizationKnowledge RepresentationSoft ComputingWeb & Text Mining
Internet address
SeriesLecture Notes in Computer Science

Bibliographical note

The conference was held virtually due to the COVID-19 pandemic.

Cite this