Traditional operating room scheduling methodologies, often reliant on large-scale linear or integer programming models, suffer from computational complexity when scaled to realistic hospital scenarios involving dozens or hundreds of surgeries per week. These conventional models require an overly high number of variables, rendering them slow and cumbersome for real-time adjustments. The Concordia-led team’s approach streamlines this by drastically reducing the dimensionality of the problem while maintaining the capability to generate near-optimal schedules. Their framework accomplishes this integration by simultaneously deciding which operating rooms to open on each day, allocating surgery start times, and identifying elective cases that may need to be deferred when emergencies arise.

The core engine of this new system is a reinforcement-learning-based column generation algorithm. Reinforcement learning, a branch of machine learning, empowers the tool to iteratively improve scheduling decisions by simulating various scenarios and learning which strategies yield the best trade-off between efficiency and flexibility. The column generation technique further optimizes the scheduling space by dynamically generating efficient scheduling patterns (columns) only when necessary, preventing the computational explosion seen in exhaustive methods. This synergy allows the model to generate high-quality operating room plans rapidly and adaptively.

One of the key innovations of this tool is its ability to perform daily schedule re-optimizations, a critical feature given the volatile nature of emergency surgical interventions. In practical hospital settings, unforeseen emergencies frequently necessitate last-minute alterations to the surgical schedule. The tool’s design acknowledges this reality by enabling the insertion of emergent cases as they arise, while minimizing the ripple effect on the existing elective procedure line-up. This balance is achieved through strategic use of limited overtime, temporary opening of additional rooms, and selective postponement of elective surgeries based on an integrated cost and disruption model.

Testing the system involved rigorous computational simulations alongside real-world application using operational data from a prominent hospital in Naples, Italy. The real-schedule data provided a challenging testbed, reflective of typical fluctuations and emergency incidences encountered in major surgical departments. Results demonstrated impressive robustness: the algorithm absorbed emergency arrivals with only marginal adjustments from the initial plan. This adaptability translated into tangible operational benefits such as fewer cancellations, optimized resource usage, and enhanced patient throughput without compromising care quality.

Beyond its immediate impact on scheduling efficiency, the AI-driven framework also holds promise in optimizing hospital operational costs. Operating rooms are among the most expensive resources in healthcare, incurring significant fixed and variable costs daily. By optimizing room utilization and reducing unnecessary cancellations, hospitals can achieve better financial performance and reduce wastage of costly surgical supplies, staff time, and specialized equipment. This holistic optimization paradigm supports not only administrative decision-makers but also clinical teams striving to meet stringent patient care targets.

The research team responsible for this innovation is composed of multidisciplinary experts in mechanical, industrial, and aerospace engineering, illustrating the increasingly cross-disciplinary nature of health technology innovation. Hossein Hashemi Doulabi, an associate professor at Concordia’s Department of Mechanical, Industrial and Aerospace Engineering, heads the project. The collaboration includes contributions from Mahdi Dolatkhah, a Concordia PhD candidate and the paper’s lead author, as well as Walter Rei from Université du Québec à Montréal and Michel Gendreau from Polytechnique Montréal, who enriched the study with their expertise in computational modeling and operations research.

The team’s paper, titled “A reinforcement-learning-based column generation algorithm for integrated operating room planning and scheduling,” was published in the International Journal of Production Research on March 13, 2026. The study publicly declares no conflicts of interest, emphasizing the integrity and objectivity of the research. Detailed insights into the algorithmic mechanisms, mathematical formulations, and validation approaches are provided within the publication, offering a valuable resource for healthcare systems scientists and operational researchers seeking to replicate or build upon this work.

Artificial intelligence and machine learning have emerged as transformative forces in healthcare, rapidly changing how complex healthcare delivery problems are approached. This research exemplifies the practical benefits of integrating AI into operational decision-making in hospitals, not merely as a theoretical tool but as an implementable solution tested in real hospital environments. It underscores the growing recognition that advanced computational methods, when thoughtfully applied, can yield measurable improvements in patient care delivery, resource management, and hospital economics.

Looking ahead, further development and deployment of this planning tool could facilitate broader adoption across healthcare systems internationally, particularly in regions facing surgical backlogs exacerbated by population growth, aging demographics, and recent global health crises. By supporting more dynamic and resilient surgical scheduling systems, hospitals could materially reduce patient wait times, improve surgical outcomes, and optimize workforce allocation. Moreover, the model’s flexibility suggests potential adaptation for other resource-constrained environments in healthcare, including diagnostic imaging departments and emergency room management.

In summary, the Concordia-led research merging reinforcement learning and column generation represents a landmark advancement in operating room scheduling technology. This integrated approach provides a pragmatic, scalable, and adaptive platform with significant implications for enhancing surgical care delivery under the common pressures of unpredictability and resource limitations. As hospitals continually seek to optimize their workflows amid emergent challenges, tools such as this could become integral to fostering healthcare system resilience and excellence in surgical care.

Subject of Research: Not applicable
Article Title: A reinforcement-learning-based column generation algorithm for integrated operating room planning and scheduling
News Publication Date: 13-Mar-2026
Web References: https://www.tandfonline.com/doi/full/10.1080/00207543.2026.2637778#d1e228
References: International Journal of Production Research, DOI: 10.1080/00207543.2026.2637778
Keywords: Health care, Health care delivery, Medical facilities, Surgery, Artificial intelligence, Machine learning

A novel security framework combining blockchain technology and reinforcement learning has been developed to tackle the unique challenges posed by IoMT systems. This cutting-edge approach offers a multidimensional layer of security designed to address not only data integrity but also privacy and real-time monitoring of potential threats. By leveraging the immutable characteristics of blockchain, data can be securely stored and reliably transmitted, reinforcing patient confidentiality. At the same time, reinforcement learning algorithms intelligently adapt to shifting threat landscapes, enhancing the overall resilience of IoMT devices against cyberattacks.

The deployment of this unique framework demonstrates a remarkable improvement in both memory consumption and transaction latency, as compared to existing traditional security methods in healthcare environments. These enhancements are critical in ensuring that healthcare providers can stream data efficiently while maintaining stringent security protocols. The findings highlight not only a heightened capability for data throughput but an improved operational efficiency, essential for seamless integration into the fast-paced world of healthcare delivery.

One of the standout features of this new approach is its effectiveness in identifying address resolution protocol man-in-the-middle attacks, achieving an accuracy rate exceeding 88%. This marks a significant advancement over traditional methods that rely on machine learning techniques, which typically report accuracies in the lower 80s. The reduction in latency—culminating in a snapshot response time of just 45 milliseconds—illustrates the framework’s potential to mitigate threats immediately, ensuring that healthcare systems remain uncompromised during critical operations.

Notably, the newly developed system also surpasses traditional machine learning by reducing false-positive rates to a mere 6%. In contrast, conventional methods often struggle with rates oscillating between 12 to 20%. The efficiency achieved by the framework allows for 80% resource utilization, although it requires distinctively higher memory usage at 320 MB. Despite these demands, the framework performed exceptionally well, especially when tested against the notorious Mirai botnet dataset, underscoring its robust capacity for evolving threat detection.

Central to this innovative framework is an advanced deep Q network based reinforcement learning model. Unlike static models, which calculate fixed responses based on predefined parameters, deep Q networks learn continuously from incoming data, thus fostering an agile response mechanism. This dynamic nature equips the framework to refine its strategies in real-time, addressing newly emerging threats with superior intelligence compared to rigid traditional models.

Hyperledger Fabric serves as the foundational blockchain technology for this system, making it an ideal component due to its efficient resource consumption and higher transaction throughput capabilities. By utilizing Hyperledger Fabric, the framework can secure validation and storage of critical data captured from IoMT sensors, thereby enhancing overall data integrity. As a result, the framework not only supports the existing cybersecurity landscape but also reinforces trust among healthcare providers and patients relying on these technologies.

While the initial results are promising, there are acknowledged limitations that necessitate further research. The current memory and computational requirements of the reinforcement learning framework are substantial, limiting its deployment in resource-constrained environments typical of many healthcare settings. Therefore, optimizing these parameters would enhance the feasibility of real-time implementation across a broader spectrum of IoMT devices.

Future developments may enhance the framework further by integrating sophisticated security measures alongside robust privacy-preserving techniques, such as federated learning. This could prove invaluable in protecting sensitive medical data shared between devices while still allowing for intelligent analytics. The exploration of hybrid models that combine the adaptability of reinforcement learning with the efficiency and simplicity of traditional algorithms could also carve pathways for innovative solutions tailored to the realities of resource-limited healthcare environments.

The urgency for robust security measures cannot be overstated given that IoMT systems are becoming increasingly commonplace in hospitals and healthcare institutions. As connected devices proliferate, so does the volume of sensitive data at risk of being compromised. Thus, frameworks that combine blockchain technology with advanced machine learning methods stand out as vital advancements in the ongoing quest to fortify security in healthcare ecosystems.

The potential for this unique framework to reshape data security in the IoMT landscape is immense, paving the way for increased patient trust and enhanced data-driven healthcare decisions. As innovation continues to evolve in this space, the adoption of secure, adaptive technologies will be paramount in ensuring that IoMT can be leveraged to its fullest potential, ultimately improving patient outcomes and amplifying the capabilities of healthcare providers worldwide.

Through these advancements, stakeholders in healthcare are presented with not just the tools to protect data but also the insights needed to harness the power of connected devices. In a world where data breaches can have unprecedented consequences, the evolution of privacy-preserving technologies is pivotal. Researchers are now poised to redefine the security landscape, promising to provide healthcare systems with solutions tailored for both efficacy and safety, where innovative technology meets the urgent demands of patient care.

Notably, the publication of this groundbreaking research reflects a broader movement among scientists and engineers who are uniting to enhance security and resilience throughout the healthcare sector. Moving forward, the integration of interdisciplinary approaches will become key to harnessing machine learning, blockchain, and IoMT technologies in synergistic ways that spell success for the future of healthcare.

Subject of Research: Not applicable
Article Title: Privacy-Preserving Strategies in the Internet of Medical Things Using Reinforcement Learning and Blockchain
News Publication Date: 14-Jul-2025
Web References: 10.34133/icomputing.0133
References: N/A
Image Credits: Dounia Doha et al.

Keywords

Applied sciences and engineering, Systems theory, Adaptive systems, Machine learning, Deep learning, Computer science, Cybersecurity