INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
Machine Learning in MRI-Based Cancer Characterization:  
Enhancing Precision and Early Detection  
Shrawan Kumar Yadav1, Dr. Harish Kumar2, Dr. Amit Kumar Janu3  
1Research Scholar, 2Guide, 3Co-Guide  
1,2 Department of Applied Sciences, NIMS University Rajasthan, Jaipur.  
3Tata Memorial Centre (ACTREC), Navi- Mumbai.  
DOI : https://doi.org/10.51583/IJLTEMAS.2025.1411000094  
Received: 03 December 2025; Accepted: 11 December 2025; Published: 22 December 2025  
ABSTRACT  
This study examines how MRI and ML can improve cancer detection and early detection. Better cancer  
diagnostics and earlier detection are the goals. Test and evaluate machine learning methods for different cancers.  
These include CNNs, SVMs, RFs, and ensemble approaches. Public datasets were analyzed. The AUC-ROC,  
sensitivity, specificity, and detection rate were used to evaluate the results. Using radiomics features with deep  
learning architectures to improve diagnostic skills is also considered. Ensemble systems that use many machine  
learning (ML) methods outperform solo algorithms with an average precision of 92.7 percent. A list of the  
primary issues that must be addressed for practical translation is also provided. This requires larger and more  
diverse datasets and improved understanding and application of learned information. Multimodal integration and  
federated learning approaches may be studied in the future to solve current issues.  
Keywords: Machine Learning, Deep Learning, MRI, Cancer Detection, Computer-Aided Diagnosis, Radiomics,  
Convolutional Neural Networks  
INTRODUCTION  
There were about 19.3 million new cases of cancer and 10 million deaths linked to cancer around the world in  
2020 (WHO, 2023) making it one of the leading causes of death in the world. For all types of cancer, the chances  
of a good treatment outcome and a high survival rate go up a lot when the disease is found early and correctly.  
Magnetic Resonance Imaging (MRI) has become a powerful, non-invasive way to diagnose and characterize  
cancer because it can show contrast in soft tissues well, can work on multiple planes, and does not give off  
ionizing radiation. However, problems with traditional MRI interpretation include the fact that different viewers  
can see things differently, it can be hard to see small details, and the imaging data is getting more and more  
complicated all the time.  
Medical image analysis has evolved due to ML and AI advances. These developments may fix these issues. MRI  
data may be diagnosed more accurately and consistently with machine learning techniques. These algorithms  
can spot complicated patterns that humans cannot. Despite advancements, utilizing machine learning in MRI to  
characterize cancer is still difficult. These issues include making the model comprehensible, compatible with  
multiple scanners and protocols, and integrating it into clinical operations.  
This research examines the newest machine learning algorithms for cancer classification utilizing MRI data from  
many body areas. Improved accuracy and earlier diagnosis are its main goals. Different machine learning  
algorithms are tested for clinical utility. Problems to solve and future research opportunities are noted. To  
demonstrate their practicality, implementation case studies for breast, prostate, and brain cancer diagnosis  
utilizing publicly available datasets are offered.  
www.ijltemas.in  
Page 971  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
BACKGROUND  
MRI in Cancer Imaging  
Magnetic resonance imaging (MRI) utilizes radio waves and strong magnetic fields in order to provide highly  
detailed anatomical and functional images. A variety of magnetic resonance imaging (MRI) sequences, including  
as T1-weighted, T2-weighted, diffusion-weighted imaging (DWI), dynamic contrast-enhanced (DCE), and  
spectroscopy, offer supplementary information regarding the features of the tissue in question.  
The advantages of MRI in cancer imaging include:  
When compared to computed tomography and ultrasound, it provides more contrast for soft tissue.  
Capabilities that are multi-parametric in nature for the purpose of functional and anatomical assessment  
Absence of ionizing radiation, permitting repeated inspections to be conducted  
Capabilities for three-dimensional imaging  
Nevertheless, there are a number of challenges that are associated with the interpretation of standard  
MRI:  
Both interobserver and intraobserver variability  
An examination of the data sets that are growing in their degree of complexity and taking a long time to  
complete  
An evaluation that is based on quality rather than quantity  
The difficulties associated with identifying minute anomalies as well as differentiating between tumors  
that are benign and those that are malignant  
Machine Learning Fundamentals  
Machine learning is a collection of computational techniques that allow systems to discover patterns within data  
without having to undergo explicit programming. When it comes to the characterization of cancer through  
magnetic resonance imaging (MRI), the different methods of machine learning can be divided into the following  
two main categories:  
Supervised learning: Typically, models are trained on labeled datasets in which the ground truth, which is usually  
confirmed through histological means, is known. This ground truth includes information such as the presence or  
absence of cancer or the classification of the condition. Support vector machines (SVMs), random forests, and  
convolutional neural networks (CNNs) are all examples of supervised learning methods that are often used.  
Learning unsupervised: These algorithms can find patterns or groupings in unlabeled data, identifying unknown  
biomarkers or patient subgroups.  
METHODOLOGY  
Literature Review Methodology  
The global health concern of prostate cancer requires better diagnosis and treatment. Prostate cancer remains the  
biggest cause of cancer, according to the ACS (2024). This prevalence highlights the need for better screening  
and management. MRI is critical for prostate cancer detection, staging, and treatment. AI, especially deep  
learning and machine learning, improves MRI prostate cancer diagnosis accuracy and efficiency. AI-based  
www.ijltemas.in  
Page 972  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
prostate cancer diagnosis and treatment utilizing magnetic resonance imaging (MRI) is reviewed in this  
literature. Recent advances in machine learning, deep learning, and emerging trends are highlighted.  
Deep learning and MRI data improved prostate cancer diagnosis and classification in a recent study. He et al.  
(2023) determine the potential of AI in MRI-based diagnosis. Li et al. (2022) analyze how machine learning has  
affected prostate MRI and recommend ways to increase accuracy and efficiency. Li et al. (2013) showed that  
MRI can characterize and stage prostate cancer tumors.  
AI can also improve MRI prostate segmentation. Salvi and colleagues (2022) segment the prostate more  
precisely using active shape models and deep learning. Alkadi et al. (2019) propose MRIbased prostate cancer  
detection and localization using deep learning. This method improves diagnosis.  
By providing accurate anatomical and functional insights, magnetic resonance imaging has changed prostate  
cancer treatment. Fernandes et al. (2022) discuss diagnosis and treatment. Turkbey et al. (2016) found that  
mpMRI can detect prostate cancer, which can be dangerous. According to Manafi-Farid et al. (2021), molecular  
imaging can improve prostate cancer diagnosis, especially in early staging.  
Combining MRI with other imaging modalities has been researched to increase diagnosis accuracy. PSMA  
PET/MRI evidences this. PSMA PET/MRI improves prostate cancer staging and therapy planning, according to  
Barbosa et al. (2018). Fischer et al. (2019) use radio genomic approaches to study tumor growth's genetic causes,  
improving customized therapy.  
Medical imaging data interpretation and trustworthiness utilizing AI have raised issues. Sadeghi et al. (2024)  
discuss healthcare openness and trust with explainable artificial intelligence (XAI). Cifci (2023) addresses  
clinical decision-making issues using an AI-based medical imaging uncertainty assessment tool.  
Deep learning in medical imaging grows. Zhu and colleagues (2024) research prostate cancer treatment with  
artificial intelligence, while Mahmood and colleagues (2023) study medical picture categorization and  
segmentation with deep learning. Research shows that AI is improving diagnosis and therapy. In addition, Sieren  
et al. (2010) highlight the technological developments in MRI and CT for cancer diagnosis, which improve AI-  
driven imaging systems.  
Using the 2005 Gleason grading system and clinical outcomes, Kryvenko and Epstein (2016) assess prostate  
cancer grading and biopsy. Heijmink et al. (2006) compares systematic and ultrasound-guided biopsies for  
diagnostic accuracy, providing a fuller picture.  
AI can improve magnetic resonance imaging diagnosis, but obstacles remain. Using MRI, SuarezIbarrola et al.  
(2022) examine the limitations and potential of AI in prostate cancer diagnosis. Almestad (2023) studies medical  
diagnoses with explainable AI and the necessity for crossdisciplinary collaboration to promote AI acceptability.  
MRI-based prostate cancer diagnostics should become more accurate, reliable, and accessible for early cancer  
detection and management after evaluating these breakthroughs and addressing present difficulties.  
Datasets  
For implementation case studies, these public datasets are used:  
Mast cancer: In the Cancer Imaging Archive's Breast-MRI-NACT-Pilot dataset, 64 histologically confirmed  
breast cancer patients underwent neoadjuvant chemotherapy. It also includes MRI scans taken both before and  
after the treatment.  
PROSTATEx Challenge Dataset: This dataset contains multi-parametric MRI data from 346 patients who have  
been diagnosed with prostate cancer. The dataset also includes the relevant histopathology findings from the  
patients' targeted biopsies.  
www.ijltemas.in  
Page 973  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
The BraTS (Brain Tumor Segmentation) 2021 dataset is comprised of multi-modal MRI scans from a total of  
369 people who have been diagnosed with gliomas of varying degrees of severity. This dataset includes T1,  
T1ce, T2, and FLAIR images.  
3.3 The Preprocessing Pipeline  
In order to standardize inputs and improve the performance of models, preprocessing MRI data is an essential  
step. The following were components of the preprocessing pipeline:  
1. Correction of the bias field to resolve the issue of intensity non-uniformity by employing the N4ITK  
algorithm  
2. The process of adjusting signal intensity so that they are consistent between and among patients  
3. Enrollment for the purpose of aligning distinct sequences in the context of multi-parametric magnetic  
resonance imaging  
Machine Learning Approaches  
Support Vector Machines (SVMs) with linear and radial basis function kernels, Random Forests with optimized  
tree depths and estimator counts, Logistic Regression utilizing L1 and L2 regularization techniques, and k-  
Nearest Neighbors (k-NN) with a range of *k* values are all examples of conventional machine learning models.  
Deep learning techniques include the use of sophisticated designs, such as convolutional neural networks  
(CNNs), which include VGG16, ResNet50, and a variety of other custom-designed models that are appropriate  
for the given task. In order to perform segmentation tasks, U-Net and its variations are employed, in addition to  
transfer learning techniques that refine the training of pre-trained networks using magnetic resonance imaging  
(MRI) datasets. In addition, the investigation of ensemble approaches that combine the results produced by a  
number of different models is undertaken in order to improve the accuracy of predictions even more.  
PyRadiomics is used to extract handcrafted characteristics, such as shape descriptors, first-order statistics, and  
texture features, in radiomic-based approaches, which are then selected using LASSO, mutual information, and  
principal component analysis.  
RESULTS  
Performance Comparison Across ML Techniques  
Table 1: Performance Comparison of ML Techniques Across Cancer Types  
ML Technique  
SVM (RBF)  
Random Forest  
CNN (VGG16)  
Ensemble  
Cancer Type  
Breast  
Accuracy (%)  
85.3  
Sensitivity (%)  
Specificity (%)  
AUC  
0.871  
0.892  
0.923  
0.941  
0.863  
0.882  
0.915  
84.1  
85.9  
89.3  
91.7  
82.5  
84.2  
88.7  
86.5  
88.4  
90.8  
93.0  
84.9  
86.8  
90.4  
Breast  
87.2  
Breast  
90.1  
Breast  
92.4  
SVM (Linear)  
Random Forest  
CNN  
Prostate  
Prostate  
Prostate  
83.7  
85.5  
89.6  
(ResNet50)  
www.ijltemas.in  
Page 974  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
Ensemble  
Prostate  
Brain  
Brain  
Brain  
Brain  
Brain  
91.8  
86.9  
88.3  
92.5  
91.4  
94.0  
90.6  
85.7  
87.1  
91.8  
90.7  
93.2  
93.0  
88.0  
89.5  
93.2  
92.0  
94.7  
0.937  
0.889  
0.905  
0.946  
0.935  
0.962  
SVM (RBF)  
Random Forest  
CNN (Custom)  
U-Net  
Ensemble  
Traditional machine learning (ML) methods were routinely outperformed by deep learning methods, with  
convolutional neural networks (CNNs) performing best across all cancer types. An ensemble method that  
blended model predictions yielded the best performance measures. This strategy improved accuracy by 2.7%  
over the best single model.  
Breast Cancer MRI Analysis Case Study  
A multi-stage pipeline used radiomics and deep learning to detect and categorize breast cancer. The model was  
trained and validated using the Breast-MRI-NACT-Pilot dataset using five-fold cross-validation.  
The ensemble model employed radiomics and CNN features to diagnose breast cancer. Radiomics factors such  
texture heterogeneity, margin sharpness, and enhancement patterns helped classify. CNN measured bodily  
qualities radiomics couldn't.  
The accuracy of diagnosing lesions under one centimeter improved by 7.6% over the radiologist. Early detection  
of cancer may benefit patients.  
Prostate Cancer Magnetic Resonance Imaging Case Study  
Prostate cancer detection in PROSTATEx dataset differentiated clinically relevant (Gleason score ≥7) from  
benign and indolent tumors.  
Figure 1: Comparison of ML models for breast cancer detection  
www.ijltemas.in  
Page 975  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
Figure 2: ROC curves for breast cancer classification models  
Figure 3: Prostate cancer detection accuracy by anatomical zone  
Figure 4: Feature importance for prostate cancer detection  
www.ijltemas.in  
Page 976  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
BraTS 2021 supported segmentation and classification. Best results were obtained with T2weighted, diffusion-  
weighted, and dynamic contrast-enhanced multi-parametric MRI. The model showed zone-specific performance  
fluctuations. It was 91.2% accurate in the perimeter and 89.7% in the transition zone. Due to benign prostatic  
hyperplasia's varied signal qualities, transition zone tumors are difficult to detect.  
The feature importance analysis showed that diffusion-weighted imaging (DWI) apparent diffusion coefficient  
(ADC) values were the most important factors in model performance, followed by DCE-MRI T2 signal intensity  
and enhancement kinetics.  
Brain Tumor Classification and Segmentation  
Case Study Utilizing Modified U-Net architecture with residual connections.  
Figure 5: Brain tumor segmentation performance by region  
Figure 6: Model complexity vs. performance for brain tumor classification  
www.ijltemas.in  
Page 977  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
In the segmentation results, the tumor volume was defined well (Dice coefficient 0.92) while the enhancing  
components were identified less well (Dice coefficient 0.85). For complex tumor margins, the ensemble  
technique dramatically increased segmentation accuracy.  
The results of the computational efficiency research showed that although larger models typically achieved better  
performance, the link between the two was not linear. Instead, the relationship diminished beyond a certain point  
of model complexity. While retaining realistic inference durations that are appropriate for clinical translation,  
the best technique balanced performance against computing limitations, attaining 93.1% classification accuracy.  
Combining Radiomic Techniques with Deep Learning Approaches  
An integrated strategy was created that incorporated standard radiomics features with deep learning techniques;  
this integrated strategy was then evaluated based on its performance in comparison to standalone methods.  
Table 2: Performance Comparison of Radiomics, Deep Learning and Hybrid Approaches  
Approach  
Radiomics  
Deep Learning  
Hybrid  
Cancer Type  
Breast  
Accuracy (%)  
84.5  
Sensitivity (%)  
83.2  
Specificity (%)  
85.8  
AUC  
0.865  
0.923  
0.948  
0.854  
0.915  
0.942  
0.873  
0.946  
0.965  
Breast  
90.1  
89.3  
90.8  
Breast  
93.2  
92.5  
93.8  
Radiomics  
Deep Learning  
Hybrid  
Prostate  
Prostate  
Prostate  
Brain  
82.9  
81.4  
84.3  
89.6  
88.7  
90.4  
92.3  
91.2  
93.4  
Radiomics  
Deep Learning  
Hybrid  
85.1  
83.9  
86.3  
Brain  
92.5  
91.8  
93.2  
Brain  
94.7  
93.9  
95.4  
When compared to solo radiomics and deep learning methods, the hybrid strategy was able to achieve better  
results across all cancer types. It was especially noticeable in situations in which the amount of labeled data was  
constrained, which implies that the information that is provided by radiomics features is extremely valuable and  
can supplement the skills of deep learning when it comes to pattern recognition.  
CONCLUSION  
This comprehensive analysis shows the huge potential of machine learning applications in magnetic resonance  
imaging (MRI)-based cancer characterization to increase cancer diagnosis accuracy and allow earlier  
identification across many cancer types. Ensemble techniques using radiomics and deep learning architectures  
work well in breast, prostate, and brain cancer cases. This strategy enhances accuracy by 5-10% over standard  
evaluation methods. This paper reports technical performance enhancements that potentially enable earlier  
intervention, more accurate characterization, and better risk classification, which could have therapeutic  
implications. Before widespread clinical implementation, model interpretability, external validation, and clinical  
process integration must be improved. Future research should focus on explainable AI technology, multimodal  
data integration, and prospective therapeutic efficacy trials. Federated learning offers novel techniques to  
manage data shortages and preserve privacy. As technology and clinical validation improve, ML-based MRI  
analysis can improve cancer care by enabling earlier detection and more precise characterization.  
www.ijltemas.in  
Page 978  
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,  
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)  
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue XI, November 2025  
REFERENCES  
1. He, M.; Cao, Y.; Chi, C.; Yang, X.; Ramin, R.; Wang, S.; Hu, K. Research progress on deep learning in  
magnetic resonance imaging–based diagnosis and treatment of prostate cancer: A review on the current  
status and perspectives. Front. Oncol. 2023, 13, 1189370.  
2. Li, H.; Lee, C.H.; Chia, D.; Lin, Z.; Huang, W.; Tan, C.H. Machine learning in prostate MRI for prostate  
cancer: Current status and future opportunities. Diagnostics 2022, 12, 289.  
3. American Cancer Society. Cancer Facts & Figures 2024. 2024. Available online:  
4. https://www.cancer.org/research/cancer-facts-statistics/all-cancer-facts-figures/2024cancer-facts-  
figures.html (accessed on 17 January 2024).  
5. Fernandes, M.C.; Yildirim, O.; Woo, S.; Vargas, H.A.; Hricak, H. The role of MRI in prostate cancer:  
Current and future directions. Magn. Reson. Mater. Phys. Biol. Med. 2022, 35, 503–521.  
6. Suarez-Ibarrola, R.; Sigle, A.; Eklund, M.; Eberli, D.; Miernik, A.; Benndorf, M.; Gratzke, C. Artificial  
intelligence in magnetic resonance imaging–based prostate cancer diagnosis: Where do we stand in  
2021? Eur. Urol. Focus 2022, 8, 409–417.  
7. Kryvenko, O.N.; Epstein, J.I. Prostate cancer grading: A decade after the 2005 modified Gleason grading  
system. Arch. Pathol. Lab. Med. 2016, 140, 1140–1152.  
8. Heijmink, S.W.; van Moerkerk, H.; Kiemeney, L.A.; Witjes, J.A.; Frauscher, F.; Barentsz, J.O. A  
comparison of the diagnostic performance of systematic versus ultrasound-guided biopsies of prostate  
cancer. Eur. Radiol. 2006, 16, 927–938.  
9. Sieren, J.C.; Ohno, Y.; Koyama, H.; Sugimura, K.; McLennan, G. Recent technological and application  
developments in computed tomography and magnetic resonance imaging for improved pulmonary  
nodule detection and lung cancer staging. J. Magn. Reson. Imaging 2010, 32, 1353–1369.  
10. Li, B.; Du, Y.; Yang, H.; Huang, Y.; Meng, J.; Xiao, D. Magnetic resonance imaging for prostate cancer  
clinical application. Chin. J. Cancer Res. 2013, 25, 240.  
11. Turkbey, B.; Brown, A.M.; Sankineni, S.; Wood, B.J.; Pinto, P.A.; Choyke, P.L. Multiparametric  
prostate magnetic resonance imaging in the evaluation of prostate cancer. CA Cancer J. Clin. 2016, 66,  
326–336.  
12. Barbosa, F.D.G.; Queiroz, M.A.; Nunes, R.F.; Marin, J.F.G.; Buchpiguel, C.A.; Cerri, G.G. Clinical  
perspectives of PSMA PET/MRI for prostate cancer. Clinics 2018, 73 (Suppl. 1), e586s.  
13. Manafi-Farid, R.; Ranjbar, S.; Jamshidi Araghi, Z.; Pilz, J.; Schweighofer-Zwink, G.; Pirich, C.;  
Beheshti, M. Molecular imaging in primary staging of prostate cancer patients: Current aspects and future  
trends. Cancers 2021, 13, 5360.  
14. Zhu, L.; Pan, J.; Mou, W.; Deng, L.; Zhu, Y.; Wang, Y.; Xue, W. Harnessing artificial intelligence for  
prostate cancer management. Cell Rep. Med. 2024, 5, 101506.  
15. Fischer, S.; Tahoun, M.; Klaan, B.; Thierfelder, K.M.; Weber, M.A.; Krause, B.J.; Hamed, M. A  
radiogenomic approach for decoding molecular mechanisms underlying tumor progression in prostate  
cancer. Cancers 2019, 11, 1293.  
16. Sadeghi, Z.; Alizadehsani, R.; CIFCI, M.A.; Kausar, S.; Rehman, R.; Mahanta, P.; Bora, P.K.; Almasri,  
A.; Alkhawaldeh, R.S.; Hussain, S.; et al. A review of Explainable Artificial Intelligence in healthcare.  
Comput. Electr. Eng. 2024, 118, 109370.  
17. Mahmood, T.; Rehman, A.; Saba, T.; Nadeem, L.; Bahaj, S.A.O. Recent advancements and future  
prospects in active deep learning for medical image segmentation and classification. IEEE Access 2023,  
11, 26143–26159.  
18. Cifci, M.A. A deep learning-based framework for uncertainty quantification in medical imaging using  
the DropWeak technique: An empirical study with baresnet. Diagnostics 2023, 13, 800.  
19. Alkadi, R.; Taher, F.; El-Baz, A.; Werghi, N. A deep learning-based approach for the detection and  
localization of prostate cancer in T2 magnetic resonance images. J. Digit.  
20. Imaging 2019, 32, 793–807.  
21. Salvi, M.; De Santi, B.; Pop, B.; Bosco, M.; Giannini, V.; Regge, D.; Meiburger, K.M. Integration of  
deep learning and active shape models for more accurate prostate segmentation in 3D MR images. J.  
Imaging 2022, 8, 133.  
22. Almestad, E. Exploring Explainable AI Adoption in Medical Diagnosis and the Empowering Potential  
of Collaboration. Master’s Thesis, NTNU, Trondheim, Norway, 2023.  
www.ijltemas.in  
Page 979