A Neural Network Approach for the Analysis of Computed Tomography Images in Adrenal Gland Diseases

A.A. Privalenko, K.S. Zaytsev, S.A. Buryakina, N.V. Tarbaeva, S.M. Kitaev, K.V. Kukhtikov

Abstract


This study investigates the potential of deep learning models for the diagnosis of adrenal gland diseases using computed tomography (CT) images. To support clinical decision-making and automate the classification of identified neoplasms in abdominal CT scans, a two-stage neural network approach, combining segmentation and classification, was developed. This approach allows for the visualization of detected lesions and accommodates multiple lesions within a single CT image. The study utilized a dataset provided by the Academician I.I. Dedov National Medical Research Center of Endocrinology, comprising 228 CT scans at the time of writing. To optimize processing time, images were converted to MP4 video format (54 frames per video), reducing data volume without significantly compromising diagnostic value. Images underwent preprocessing, and data augmentation was employed to address class imbalance. Each CT scan was annotated with three labels, corresponding to the presence of a neoplasm with a malignant, benign, or indeterminate phenotype. For lesion instance segmentation, a YOLOv11-seg model pre-trained on the COCO dataset was implemented. A 3DResNet-50 model, trained on the segmented regions, was used for classification. The proposed combined two-stage approach is implemented in a software suite designated “Assistant Endocrinologist”.

Full Text:

PDF (Russian)

References


Tobias Carling «How is adrenal cancer diagnosed?» [Electronic resource]. Available: https://www.adrenal.com/blog/how-is-adrenal-cancer-diagnosed.

Andreeva A.V. Podhod vracha obshhej praktiki k vedeniju pacientov s incidentalomami nadpochechnikov. Terapevt, 2019, # 12 s. 70-78.

Adrenal incidentalomas: imaging challenges—role of MDCT scan versus MRI in evaluating adrenal incidentalomas. [Electronic resource]. Available: https://ejrnm.springeropen.com/articles/10.1186/s43055-021-00437-w.

U-Net: Convolutional Networks for Biomedical Image Segmentation. [Electronic resource]. Available: https://link.springer.com/chapter/10.1007/978-3-319-24574-4_28.

Y-Net: Joint Segmentation and Classification for Diagnosis of Breast Biopsy Images [Electronic resource]. Available: https://link.springer.com/chapter/10.1007/978-3-030-00934-2_99.

Hybrid Deep Learning Framework for Classification of Kidney CT. [Electronic resource]. Available: arXiv:2502.04367v1.

Accurate classification of lung nodules on CT images using the TransUnet. [Electronic resource]. Available: https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.1060798/full.

Using CT radiomic features based on machine learning models to subtype adrenal adenoma. [Electronic resource]. Available: https://link.springer.com/article/10.1186/s12885-023-10562-6.

Computed Tomography-Based Machine Learning Differentiates Adrenal Pheochromocytoma From Lipid-Poor Adenoma. [Electronic resource]. Available: https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2022.833413/full?utm_source=chatgpt.com.

Machine learning-based quantitative texture analysis of CT images of small renal masses: Differentiation of angiomyolipoma without visible fat from renal cell carcinoma. [Electronic resource]. Available: https://link.springer.com/article/10.1007/s00330-017-5118-z?fromPaywallRec=true.

Machine learning-based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans. [Electronic resource]. Available: https://link.springer.com/article/10.1007/s00261-021-03136-2.

GitHub - rafiibnsultan/3D-CT-Image-Segmentation-Using-MONAI. [Electronic resource]. Available: https://github.com/rafiibnsultan/3D-CT-Image-Segmentation-Using-MONAI.

Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT. [В Интернете]. Available: https://pubs.rsna.org/doi/full/10.1148/radiol.220101.

Diagnostic Accuracy of CT Texture Analysis in Adrenal Masses: A Systematic Review. [Electronic resource]. Available: https://www.mdpi.com/1422-0067/23/2/637.

Video Pretraining Advances 3D Deep Learning on Chest CT Tasks. [Electronic resource]. Available: https://ar5iv.labs.arxiv.org/html/2304.00546.

EfficientNet family U-Net models for deep learning semantic segmentation of kidney tumors on CT images. [Electronic resource]. Available: https://www.frontiersin.org/articles/10.3389/fcomp.2023.1235622/full.

Multi-Scale Supervised 3D U-Net for Kidneys and Kidney Tumor Segmentation. [Electronic resource]. Available: https://arxiv.org/abs/2004.08108.

FPN-SE-ResNet Model for Accurate Diagnosis of Kidney Tumors Using CT Images. [Electronic resource]. Available: https://www.mdpi.com/2076-3417/13/17/9802.

Liver Tumor Localization Based on YOLOv3 and 3D-Semantic Segmentation Using Deep Neural Networks. [Electronic resource]. Available: https://pubmed.ncbi.nlm.nih.gov/35453870/.

An attempt at beating the 3D U-Ne. [Electronic resource]. Available: https://arxiv.org/abs/1908.02182.

UniverSeg: Universal Medical Image Segmentation. [Electronic resource]. Available: https://vitalab.github.io/article/2024/05/24/universeg.html.

Automated Adrenal Gland Disease Classes Using Patch-Based Center Symmetric Local Binary Pattern Technique with CT Images. [Electronic resource]. Available: https://link.springer.com/article/10.1007/s10278-022-00759-9?fromPaywallRec=true.

CT-based deep learning model for the prediction of DNA mismatch repair deficient colorectal cancer: a diagnostic study. [Electronic resource]. Available: https://pubmed.ncbi.nlm.nih.gov/36949511/.

A transfer learning based deep learning model to diagnose covid-19 CT scan images. [Electronic resource]. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC9177227/?utm_source=chatgpt.com.

Detecting adrenal lesions on 3D CT scans using a 2.5D deep learning model. [Electronic resource]. Available: https://www.medrxiv.org/content/10.1101/2023.02.22.23286184v1.

3D Deep Learning from CT Scans Predicts Tumor Invasiveness of Subcentimeter Pulmonary Adenocarcinomas. [Electronic resource]. Available: https://aacrjournals.org/cancerres/article/78/24/6881/633067/3D-Deep-Learning-from-CT-Scans-Predicts-Tumor.

Boosting Learning by Representing 3D Medical Imaging to 2D Features for Small Data. [Electronic resource]. Available: https://arxiv.org/html/2002.04251v3?.

3D Deep Learning on Medical Images: A Review. [Electronic resource]. Available: https://arxiv.org/abs/2004.00218.

Ultralytics YOLO11,» [Electronic resource]. Available: https://docs.ultralytics.com/models/yolo11/.

3D radiotherapy dose prediction on head and neck cancer patients with a hierarchically densely connected U-net deep learning architecture. [Electronic resource]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/ab039b/meta.

ViPTT-Net: Video pretraining of spatio-temporal model for tuberculosis type classification from chest CT scans. [Electronic resource]. Available: https://arxiv.org/abs/2105.12810.

Vision Transformers: State of the Art and Research Challenges. [Electronic resource]. Available: https://arxiv.org/abs/2207.03041.


Refbacks

  • There are currently no refbacks.


Abava  Кибербезопасность ИБП для ЦОД СНЭ

ISSN: 2307-8162