Artificial Intelligence for Intelligent and Privacy-Preserving Healthcare: A Survey of Medical Image Analysis, Multimodal Diagnosis, Semi-Supervised Learning, and Personalized Healthcare Systems

Abstract

Artificial intelligence (AI) is increasingly embedded in healthcare, from image-based diagnosis to patient-facing management tools, yet the supporting literature is fragmented across communities that rarely read one another. Motivated by the need for a coherent view, this survey examines four recent peer-reviewed studies that together span multimodal COVID-19 diagnosis, privacy-preserving learning on encrypted medical images, graph-refined semi-supervised pathology classification, and an AI-enabled mobile application for personalized healthcare management and doctor recommendation. Using a structured narrative methodology, the reviewed studies are classified into a five-category taxonomy, compared along technical dimensions, healthcare capabilities, and reported limitations, and synthesized into a layered view that treats imaging, multimodal fusion, privacy preservation, label-efficient learning, and personalization as complementary layers of intelligent healthcare rather than independent technologies. The synthesis indicates that each study advances one layer under its own experimental setting, and that none addresses privacy, label efficiency, multimodality, explainability, and clinical validation together. Recurring gaps include limited clinical validation, dataset constraints, unquantified generalization across institutions, privacy–performance trade-offs, and inconsistent evaluation protocols that prevent direct numerical comparison. Future directions include privacy-preserving multimodal learning, reduced reliance on annotations, explainable and robust models, integration of imaging with clinical records, and multi-site clinical validation. This review is not exhaustive, and no new experiments are presented.

Introduction

The text presents a survey of artificial intelligence (AI) and machine learning (ML) applications in healthcare, focusing on how these technologies improve medical diagnosis, protect sensitive medical data, reduce the need for labeled datasets, and support personalized healthcare services.

The survey examines four major studies representing different areas of intelligent healthcare. The first study uses multimodal learning to combine medical imaging and laboratory information, such as chest X-rays, CT scans, and RT-PCR results, to improve COVID-19 detection. The second explores privacy-preserving AI through learnable image encryption, allowing deep learning models to analyze protected medical images without directly exposing their original content. The third applies semi-supervised learning and graph-based segmentation to whole-slide pathology images, refining pseudo-labels to reduce dependence on expensive manual annotations. The fourth presents a personalized healthcare application that provides symptom-based specialist recommendations, diet and exercise planning, prescription management, and electronic health record access.

The survey explains several foundational technologies used in healthcare AI, including convolutional neural networks (CNNs), ResNet-50, DenseNet-121, XceptionNet, natural language processing (NLP), and graph-based learning. It also highlights the importance of multimodal data integration, privacy protection, efficient learning from limited labeled data, and personalized clinical decision support.

The methodology follows a structured narrative or scoping approach. Four peer-reviewed IEEE papers supplied by the authors form the primary evidence base, while additional references are drawn from those papers. The studies are compared according to their objectives, models, datasets, evaluation methods, reported results, limitations, and future research directions. The survey does not conduct an independent systematic database search, perform a meta-analysis, or present new experiments.

Despite the progress demonstrated by the reviewed studies, several challenges remain. These include the high computational requirements of deep learning, limited availability of annotated medical data, privacy and security risks, unreliable pseudo-labels, limited model explainability, and difficulties in generalizing results across different patient populations and clinical environments. Personalized healthcare applications must also ensure data security and provide appropriate support without replacing professional medical judgment.

In conclusion, the survey organizes intelligent healthcare research into five interconnected areas: medical image analysis and deep learning, multimodal diagnosis, privacy-preserving AI, semi-supervised and graph-based learning, and personalized healthcare systems. It emphasizes that integrating these approaches could improve diagnostic reliability, data privacy, resource efficiency, and patient-centered care. Future research should prioritize explainable AI, robust validation, secure data handling, improved generalization, and practical clinical deployment to make intelligent healthcare systems more reliable and effective.

Conclusion

AI is used across healthcare, from image interpretation to patient-facing management. The reviewed studies show that medical image analysis remains central, that fusing complementary modalities improved results in one setting [1], that privacy-preserving training is feasible but trades accuracy for protection [2], that graph-refined pseudo-labels can reduce annotation needs [3], and that personalized applications can integrate recommendation, scheduling, and record access [4]. Shared challenges include limited clinical validation, small or narrow datasets, unquantified generalization, privacy–performance trade-offs, limited explainability, and non-comparable evaluation. Future healthcare AI will need to be reliable, secure, explainable, and clinically validated, and combining these layers remains an open research problem.

References

Items [1]–[4] are the seed studies. Items [5]–[28] are secondary references cited within the seed papers; their details are reproduced as given there and were not independently verified against a publisher database. [1] A. M. Arul Raj and R. Sugumar, \"Multi-modal fusion of deep learning with CNN based COVID-19 detection and classification combining chest X-ray images,\" in Proc. 7th Int. Conf. Intell. Comput. Control Syst. (ICICCS), 2023, pp. 569–575, doi: 10.1109/ICICCS56967.2023.10142910. [2] Q.-X. Huang, W. L. Yap, M.-Y. Chiu, and H.-M. Sun, \"Privacy-preserving deep learning with learnable image encryption on medical images,\" IEEE Access, vol. 10, pp. 66345–66355, 2022, doi: 10.1109/ACCESS.2022.3185206. [3] H.-K. Shin, K.-H. Uhmn, K. Choi, Z. Xu, S.-W. Jung, and S.-J. Ko, \"Graph segmentation-based pseudo-labeling for semi-supervised pathology image classification,\" IEEE Access, vol. 10, pp. 93960–93970, 2022, doi: 10.1109/ACCESS.2022.3204000. [4] A. N. Saritha, G. Kalnoor, P. R. Bhat, P. A. Rao, P. Prajwal, and P. T. Pradeep, \"Alleviate: An AI-powered mobile application for personalized healthcare management and doctor recommendation,\" IEEE Access, vol. 13, pp. 215822–215832, 2025, doi: 10.1109/ACCESS.2025.3647089. [5] C. L. Srinidhi, O. Ciga, and A. L. Martel, \"Deep neural network models for computational histopathology: A survey,\" Med. Image Anal., vol. 67, Jan. 2021, Art. no. 101813. [6] G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, \"Densely connected convolutional networks,\" in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jul. 2017, pp. 4700–4708. [7] F. Chollet, \"Xception: Deep learning with depthwise separable convolutions,\" in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jul. 2017, pp. 1251–1258. [8] F. Hao, X. Liu, M. Li, and W. Han, \"Accurate kidney pathological image classification method based on deep learning and multi-modal fusion method with application to membranous nephropathy,\" Life, vol. 13, no. 2, Jan. 2023, doi: 10.3390/life13020399. [9] M. Tanaka, \"Learnable image encryption,\" in Proc. IEEE Int. Conf. Consum. Electron.-Taiwan (ICCE-TW), May 2018, pp. 1–2. [10] W. Sirichotedumrong, Y. Kinoshita, and H. Kiya, \"Pixel-based image encryption without key management for privacy-preserving deep neural networks,\" IEEE Access, vol. 7, pp. 177844–177855, 2019. [11] J. E. Van Engelen and H. H. Hoos, \"A survey on semi-supervised learning,\" Mach. Learn., vol. 109, no. 2, pp. 373–440, 2020. [12] D.-H. Lee, \"Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks,\" in Proc. Workshop Challenges Represent. Learn., 2013, p. 896. [13] E. Arazo, D. Ortego, P. Albert, N. E. O’Connor, and K. McGuinness, \"Pseudo-labeling and confirmation bias in deep semi-supervised learning,\" in Proc. Int. Joint Conf. Neural Netw. (IJCNN), Jul. 2020, pp. 1–8. [14] M. N. Rizve, K. Duarte, Y. S. Rawat, and M. Shah, \"In defense of pseudo-labeling: An uncertainty-aware pseudo-label selection framework for semi-supervised learning,\" in Proc. Int. Conf. Learn. Represent., 2020. [15] Y. Boykov and G. Funka-Lea, \"Graph cuts and efficient N-D image segmentation,\" Int. J. Comput. Vis., vol. 70, no. 2, pp. 109–131, Nov. 2006. [16] S.-K. Tan, S.-C. Chong, K.-K. Wee, and L.-Y. Chong, \"Personalized healthcare: A comprehensive approach for symptom diagnosis and hospital recommendations using AI and location services,\" J. Informat. Web Eng., vol. 3, no. 1, pp. 117–135, Feb. 2024. [17] J. Tseng et al., \"An interactive healthcare system with personalized diet and exercise guideline recommendation,\" in Proc. Conf. Technol. Appl. Artif. Intell. (TAAI), 2015, pp. 525–532. [18] A. H. Chang and B. M. Case, \"Attacks on image encryption schemes for privacy-preserving deep neural networks,\" 2020, arXiv:2004.13263. [19] W. Sirichotedumrong and H. Kiya, \"A GAN-based image transformation scheme for privacy-preserving deep neural networks,\" in Proc. 28th Eur. Signal Process. Conf. (EUSIPCO), Jan. 2021, pp. 745–749. [20] Y. Ding et al., \"DeepEDN: A deep-learning-based image encryption and decryption network for internet of medical things,\" IEEE Internet Things J., vol. 8, no. 3, pp. 1504–1518, Feb. 2021, doi: 10.1109/JIOT.2020.3012452. [21] R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, \"Grad-CAM: Visual explanations from deep networks via gradient-based localization,\" in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Oct. 2017, pp. 618–626. [22] B. E. Bejnordi et al., \"Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer,\" JAMA, vol. 318, no. 22, pp. 2199–2210, Dec. 2017. [23] N. A. Koohbanani, B. Unnikrishnan, S. A. Khurram, P. Krishnaswamy, and N. Rajpoot, \"Self-Path: Self-supervision for classification of pathology images with limited annotations,\" IEEE Trans. Med. Imag., vol. 40, no. 10, pp. 2845–2856, Oct. 2021. [24] J. Silva-Rodriguez, A. Colomer, J. Dolz, and V. Naranjo, \"Self-learning for weakly supervised Gleason grading of local patterns,\" IEEE J. Biomed. Health Informat., vol. 25, no. 8, pp. 3094–3104, Aug. 2021. [25] L. Björkdahl, O. Pauli, J. Östman, C. Ceccobello, S. Lundell, and M. Kjellberg, \"Towards holistic disease risk prediction using small language models,\" 2024, arXiv:2408.06943. [26] M. M. Rahman et al., \"Web-based multiple disease prediction system with transfer learning and healthcare chatbot,\" in Proc. IEEE 3rd Int. Conf. Comput. Mach. Intell. (ICMI), Jan. 2024, pp. 1–5. [27] R. Surve, T. Purohit, R. Joseph, and P. Shaikh, \"HealthCare chatbot using machine learning and NLP,\" in Proc. 6th Int. Conf. Adv. Sci. Technol. (ICAST), Dec. 2023, pp. 411–416. [28] R. Umbare, R. Patil, T. Mukund, A. Rathod, and P. Rajurkar, \"A smart healthcare application using artificial intelligence (AI) and machine learning (ML),\" in Proc. 5th Int. Conf. Image Process. Capsule Netw. (ICIPCN), Jul. 2024, pp. 204–210.

Copyright

Copyright © 2026 Abhinav Krishna, Akshay G, Abhishek Sreeji, Abiya F R, Mrs. Sagara M R. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.