Applying Convolutional Neural Network for Cancer Disease Diagnosis Based on Gene Expression Data

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F44555601%3A13440%2F23%3A43898152" target="_blank" >RIV/44555601:13440/23:43898152 - isvavai.cz</a>

Result on the web

<a href="https://www.scopus.com/record/display.uri?eid=2-s2.0-85182265720&origin=resultslist" target="_blank" >https://www.scopus.com/record/display.uri?eid=2-s2.0-85182265720&origin=resultslist</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Applying Convolutional Neural Network for Cancer Disease Diagnosis Based on Gene Expression Data

Original language description

Applying deep learning techniques, such as convolutional or recurrent neural networks, to process gene expression data for developing complex disease diagnostic systems is one of modern bioinformatics&apos;s current focuses. Deep learning algorithms can identify specific patterns in the hierarchical representation of data and craft distinct functions that allow for precise identification of the subjects being studied. In this paper, we present our research findings on applying a convolutional neural network (CNN) in diagnosing various types of cancer based on gene expression data. The experimental data were sourced from The Cancer Genome Atlas (TCGA) and comprised 3269 samples. These samples can be categorized into nine classes based on the type of cancer. We introduced an ordered search-by-grid algorithm to pinpoint the optimal set of hyperparameters for the CNN. We assessed the model&apos;s efficacy using classification quality metrics, considering type I and II errors. Furthermore, we introduced an integrated F1-score index, drawing from the Harrington desirability function. The obtained results demonstrate the high efficacy of our proposed approach in diagnosing cancer based on gene expression data. The simulation results have shown that the single-layer CNN is more efficient for this type of data by all classification quality criteria. The number of correctly identified samples was 955 out of 981. The classification accuracy was 97.3%.

Czech name

—

Czech description

—

Type

W - Workshop organization

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

—

Continuities

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2023

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Event location

Bratislava

Event country

SK - SLOVAKIA

Event starting date

—

Event ending date

—

Total number of attendees

120

Foreign attendee count

85

Type of event by attendee nationality

WRD - Celosvětová akce