In healthcare AI, the most consequential advances are often those that bridge the gap between algorithmic power and clinical trust. As machine learning models grow more sophisticated, the question of interpretability has become not merely academic but operational-a prerequisite for adoption in environments where decisions directly affect human lives. Sudeep Mungara’s contributions to this space reflect a practitioner’s understanding of this challenge. His work spans systematic research on explainable artificial intelligence, documentation of emerging surgical technologies, and the conceptualization of neurological monitoring systems designed for continuous, accessible care. Across these efforts runs a consistent thread: making AI systems transparent, interpretable, and deployable in real clinical contexts.
Explainable AI in disease management
Healthcare practitioners have long expressed preference for transparent models over marginally more accurate “black box” systems. This reality-that clinical adoption depends on interpretability as much as performance-forms the foundation of Mungara’s systematic literature review published in the International Journal of Soft Computing and Engineering. “Healthcare Through AI: Integrating Deep Learning, Federated Learning, and XAI for Disease Management” examines how explainable AI techniques are being integrated across the machine learning spectrum, from traditional supervised learning to federated architectures that enable collaborative model training while preserving patient privacy. The study synthesizes findings from 24 peer-reviewed papers (screened from an initial corpus of 419 publications), revealing an acceleration in XAI research-with 38% of relevant studies published in 2022 alone. What distinguishes this work is its systematic identification of gaps in the field. While 14 of the 24 studies reviewed employed traditional machine learning with XAI overlays, only 3 addressed federated learning-despite its clear advantages for privacy-preserving medical AI. The research identifies heart disease, Parkinson’s disease, and respiratory conditions as primary domains where XAI-enabled systems are being validated, providing researchers and practitioners with a structured view of where the field stands and where opportunities remain. The methodology itself-moving from 419 publications through multiple screening stages to 24 highly relevant studies-demonstrates the kind of rigorous filtering necessary to separate signal from noise in a rapidly expanding research domain.
Surgical intelligence at the technology frontier
If the XAI work maps the interpretability landscape, Mungara’s contribution to surgical data science examines how AI, computer vision, and mixed reality converge in operating room environments. Published in the International Journal of Preventive Medicine and Health, “Surgical Data Science and Associated Techniques Facilitate the Development of Contemporary Equipment like Apple’s Vision Pro” provides one of the first comprehensive technical assessments of spatial computing technology in clinical practice. The paper documents Apple’s Vision Pro not as consumer technology but as a clinical instrument: 22 million pixels of resolution, 100Hz refresh rate, 6 degrees of freedom motion detection. More importantly, it reports validation metrics that matter for surgical precision-1.3mm mean error in 3D target visualization, with 97% of craniotomy trajectory paths falling within a 1.5mm margin. These are the thresholds that determine whether augmented reality systems can support neurosurgical decision-making or remain demonstration technology. The work maps how surgical data science-the capture, organization, and modeling of surgical workflow data-creates the foundation for these technologies. Natural language processing extracts insights from clinical records. Convolutional neural networks analyze intraoperative imaging. Real-time decision support systems provide procedural guidance. The paper documents clinical pilots across neurosurgery (spinal arteriovenous fistula treatment), plastic surgery education, and ophthalmology screening, moving beyond concept demonstrations to documented medical applications. Importantly, the research addresses current limitations candidly: the Vision Pro’s mixed-reality pass-through resolution, while sufficient for data overlay, currently restricts highly intricate manual tasks. This assessment reflects an engineering perspective-focused not on promotional claims but on operational readiness.
Conceptualizing continuous neurological monitoring
Beyond published research, Mungara’s patent application for an “AI Enabled Device for Detection of Neurological Disorders” represents an effort to translate research principles into deployable medical technology. The conceptualized system integrates wearable EEG sensors with cloud-based AI processing to enable continuous neurological monitoring outside clinical facilities. Unlike traditional diagnostic methods-MRI and CT scans that are expensive, facility-bound, and episodic-the proposed approach uses non-invasive headgear with multi-channel EEG sensor arrays, onboard preprocessing to filter artifacts, and convolutional and recurrent neural networks for pattern recognition. A mobile application interface provides real-time access for both patients and providers, while cloud storage enables longitudinal tracking. The innovation addresses several practical constraints: it shifts neurological screening from imaging suites to home environments; enables ongoing assessment beyond traditional episodic testing; and targets significantly lower costs than traditional neuroimaging. The focus on early detection positions the system for presymptomatic identification, when interventions are most effective. The patent documentation addresses not just algorithm architecture but practical deployment considerations including data security and system integration. These details reflect engagement with the operational realities that determine whether medical technology concepts become clinical tools. Three principles emerge across this body of work. First, transparency as a design requirement-both papers emphasize that clinical AI must be interpretable, whether through explicit XAI techniques or through augmented reality interfaces that make AI reasoning visible to clinicians in real time. Second, privacy-preserving architecture-the XAI paper’s attention to federated learning reflects growing recognition that powerful AI and strict privacy protection must coexist in healthcare. The patent work similarly emphasizes secure data handling and controlled access. Third, validation rigor-rather than claiming breakthroughs, the research documents measured progress with quantified metrics, acknowledges current limitations, and identifies areas requiring further development. This approach is characteristic of engineering research grounded in practical constraints rather than speculative possibilities.
Healthcare AI systems
Mungara’s professional work complements these research contributions through direct involvement in healthcare AI system development. His technical expertise encompasses machine learning model development, natural language processing, computer vision, and the architectural challenges of deploying AI in production healthcare environments-experience that informs the practical orientation evident in his research and patent work. This combination of systematic research capability, technical documentation skill, and production AI experience reflects the kind of cross-disciplinary expertise increasingly necessary as healthcare AI moves from research prototype to clinical deployment. The challenges are no longer purely algorithmic; they involve interpretability frameworks, regulatory compliance, privacy architectures, and clinical workflow integration.
Research to practice
As AI systems transition from academic papers to clinical tools, the work represented here-mapping the XAI landscape, documenting emerging surgical technologies, conceptualizing accessible monitoring systems-addresses different facets of this translation challenge. The XAI research provides systematic understanding of how interpretability is being achieved across different domains. The surgical data science work documents how spatial computing and AI convergence is performing in actual clinical settings. The patent conceptualizes how continuous monitoring might overcome accessibility barriers in neurological care. Whether this particular body of work proves influential will depend on factors that unfold over years: whether other researchers build on these systematic frameworks, whether the clinical use cases documented here scale to broader adoption, whether the monitoring concepts navigate regulatory approval and clinical validation. What is evident now is sustained engagement with the operational challenges of medical AI deployment-interpretability, privacy, accuracy validation, workflow integration, and regulatory readiness. These are the challenges that determine which AI systems remain in journals and which become clinical tools. In healthcare technology, that orientation toward practical deployment challenges often distinguishes research that shapes practice from research that documents possibilities.
Adrian Kowalsky is a technology reporter known for her incisive analysis and coverage of the tech sector. With a background in data science from Stanford University, Adrian shifted to journalism to explore the intersection of technology and society. He currently writes news and opinions focusing on tech, startups, privacy issues, and innovation trends.
