Tamás Katona

Dr. Tamás Katona

Faculty of Informatics, University of Debrecen, Debrecen, Hajdú-Bihar, Hungary
orcid

Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
8 March 2024 in Mathematics
Verified
Journal Article
Open access
PDF file
Tamás KatonaTamás Katona
Tamás Katona
GTGábor Tóth
Gábor Tóth
MPMátyás Petró
Mátyás Petró
Balazs HarangiBalazs Harangi
Balazs Harangi
Chest X-ray evaluation is challenging due to its high demand and the complexity of diagnoses. In this study, we propose an optimized deep learning model for the multi-label classification of chest X-ray images. We leverage pretrained convolutional neural networks (CNNs) such as VGG16, ResNet 50, and DenseNet 121, modifying their output layers and fine-tuning the models. We employ a novel optimization strategy using the Hyperband algorithm to efficiently search the hyperparameter space while adjusting the fully connected layers of the CNNs. The effectiveness of our approach is evaluated on the basis of the Area Under the Receiver Operating Characteristic Curve (AUC-ROC) metric. Our proposed methodology could assist in automated chest radiograph interpretation, offering a valuable tool that can be used by clinicians in the future.

6 July 2025 in Diagnostics
Verified
Journal Article
Open access
PDF file
SCSzilvia Csehely
Szilvia Csehely
AKAdrienn Kun
Adrienn Kun
EOEdina Orbán
Edina Orbán
Tamás KatonaTamás Katona
Tamás Katona
MOMonika Orosz
Monika Orosz
AJAttila Jakab
Attila Jakab
Background: Premature ovarian insufficiency (POI) is a complex and heterogeneous condition affecting women of reproductive age. Historically, most POI cases have been classified as idiopathic due to limited diagnostic capabilities. However, due to the success of oncologic treatments and the increasing number of gynecologic surgeries enabled by improved diagnostics, the proportion of iatrogenic POI cases has risen substantially. Objectives: To investigate the current prevalence of POI etiologies, to compare the etiological distribution between two POI cohorts from a single tertiary center—one historical (1978–2003) and one contemporary (2017–2024)—and to explore how the spectrum of underlying causes has changed over the past four decades. Methods: Data from 111 women diagnosed with POI between 2017 and 2024 were retrospectively reviewed and compared with those from a historical cohort of 172 patients. Etiologies were classified as genetic, autoimmune, iatrogenic, or idiopathic. Statistical comparisons were performed using chi-square and z-tests. Hormonal profiles and reproductive outcomes were also analyzed. Results: The current prevalence of POI etiologies is as follows: genetic 9.9%, autoimmune 18.9%, iatrogenic 34.2%, idiopathic 36.9%. In the historical POI cohort, etiologies were classified as genetic in 11.6%, autoimmune in 8.7%, iatrogenic in 7.6%, and idiopathic in 72.1%. The changes in the prevalence of autoimmune, iatrogenic, and idiopathic POI were statistically significant (p < 0.05). Reproductive outcomes remained limited: 10 pregnancies occurred in each cohort, with 7 live births in the contemporary group. Conclusions: Our findings suggest a significant shift in the etiological landscape of POI, with a notable, more than fourfold rise in identifiable iatrogenic cases and a twofold increase in the autoimmune group, resulting in a halving of idiopathic POI. Prevalence of genetic etiology remained unchanged. While diagnostic capabilities have improved, reproductive outcomes remain largely unchanged and suboptimal.

Verified
Journal Article
Open access
PDF file
Tamás KatonaTamás Katona
Tamás Katona
GTGábor Tóth
Gábor Tóth
MPMátyás Petró
Mátyás Petró
Balazs HarangiBalazs Harangi
Balazs Harangi
Chest X-rays are vital in healthcare for diagnosing various conditions due to their low Radiation exposure, widespread availability, and rapid interpretation. However, their interpretation requires specialized expertise, which can limit scalability and delay diagnoses. This study addresses the multi-label classification challenge of chest X-ray images using the Chest X-ray14 dataset. We propose a novel online ensemble technique that differs from previous penalty-based methods by focusing on combining individual model losses with the overall ensemble loss. This approach enhances interaction and feedback among models during training. Our method integrates multiple pre-trained CNNs using strategies like combining CNNs through an additional fully connected layer and employing a label-weighted average for outputs. This multi-layered approach leverages the strengths of each model component, improving classification accuracy and generalization. By focusing solely on image data, our ensemble model addresses the challenges posed by null vectors and diverse pathologies, advancing computer-aided radiology.

Verified
Journal Article
Open access
PDF file
SCSzilvia Csehely
Szilvia Csehely
AKAdrienn Kun
Adrienn Kun
EOEdina Orbán
Edina Orbán
Tamás KatonaTamás Katona
Tamás Katona
MOMonika Orosz
Monika Orosz
AJAttila Jakab
Attila Jakab
Background/Objectives: Premature ovarian insufficiency (POI) is a leading cause of hypoestrogenism in women under the age of 40 years and is associated with an increased risk of impaired bone health. Early diagnosis and timely hormonal intervention are essential to prevent irreversible bone loss. However, diagnostic delay is not uncommon in clinical practice. Methods: We conducted a retrospective analysis of 168 women diagnosed with POI or early menopause (EM) between 2017 and 2024 at a tertiary gynecological endocrinology unit. Bone mineral density (BMD) and T-score were assessed by dual-energy X-ray absorptiometry (DXA) at the time of diagnosis in 125 patients, of whom 116 had secondary amenorrhea. The interval between the last menstrual period (LMP) and diagnosis was used to assess the impact of diagnostic delay. The patients were further stratified by serum estradiol (E2) levels and body mass index (BMI). Results: At the time of diagnosis, 43.1% of patients had osteopenia, and 10.3% had osteoporosis. A statistically significant negative correlation was observed between time to diagnosis and BMD (r = −0.225, p = 0.022), with a similar trend seen for T-score (r = −0.211, p = 0.031). In patients with E2 ≤ 5 ng/L, the association was stronger (BMD: r = −0.401, p = 0.026). Lower E2 levels tended to be associated with poorer bone health in women with a BMI < 25 kg/m2, whereas no such trend was observed in those with a higher BMI. Conclusions: Our findings indicate that diagnostic delay in POI is associated with deterioration in bone health, particularly in lean patients and those with severe hypoestrogenism. These results underscore the importance of early recognition and timely initiation of hormone therapy to preserve bone mass and reduce long-term skeletal complications.

5 August 2022 in Cytometry Part A
Verified
Journal Article
Open access
IRIstván Rebenku
István Rebenku
FBFerenc A. Bartha
Ferenc A. Bartha
Tamás KatonaTamás Katona
Tamás Katona
BZBarbara Zsebik
Barbara Zsebik
GAGéza Antalffy
Géza Antalffy
GVGyörgy Vereb
György Vereb
The emergence and fast advance of digital pathology allows the acquisition, digital storage, interactive recall and analysis of morphology at the tissue level. When applying immunohistochemistry, it also affords the correlation of morphology with the expression of one or two specific molecule of interest. The rise of fluorescence pathology scanners expands the number of detected molecules based on multiplex labeling. The Pannoramic Confocal (created by 3DHistech, Hungary) is a first‐of‐the‐kind digital pathology scanner that affords not only multiplexed fluorescent detection on top of conventional transmission imaging, but also confocality. We have benchmarked this scanner in terms of stability, precision, light efficiency, linearity and sensitivity. X–Y stability and relocalisation precision were well below resolution limit (≤50 nm). Light throughput in confocal mode was 4–5 times higher than that of a point scanning confocal microscope, yielding similar calculated confocal intensities but with the potential for improving signal to noise ratio or scan speed. Response was linear with R2 ≥ 0.9996. Calibrated measurements showed that using indirect labeling ≥2000 molecules per cell could be well detected and imaged on the cell surface. Both standard‐based and statistical post‐acquisition flatfield corrections are implemented. We have also measured the point spread function (PSF) of the instrument. The dimensions of the PSF are somewhat larger and less symmetric than of the theoretical PSF of a conventional CLSM, however, the spatial homogeneity of these parameters allows for obtaining a specific system PSF for each optical path and using it for optional on‐the‐fly deconvolution. In conclusion, the Pannoramic Confocal provides sensitive, quantitative widefield and confocal detection of multiplexed fluorescence signals, with optical sectioning and 3D reconstruction, in addition to brightfield transmission imaging. High speed scanning of large samples, analysis of tissue heterogeneity, and detection of rare events open up new ways for quantitatively analyzing tissue sections, organoid cultures or large numbers of adherent cells.

1 September 2019
Verified
Conference Paper
Closed access
Tamás KatonaTamás Katona
Tamás Katona
BABalint Antal
Balint Antal
Radiography has outmost importance in recognizing several life threatening diseases. Nowadays, the X-ray machines can produce more examinations on a day what the physicians can analyze. In this paper, we propose a light-weight, automated, scalable method to analyze X-ray images. We designed and trained a Deep Convolutional Neural Network to recognize the presence of 16 medical conditions. We evaluated our method on publicaly available data and showed that our method is on par with state-of-the-art but has much lower computational burden.