Pleural mesothelioma poses unique diagnostic challenges because it develops as a thin, irregular layer along the lining of the lungs rather than forming discrete masses. This morphology renders existing international standards like the RECIST (Response Evaluation Criteria in Solid Tumors) inadequate. RECIST depends primarily on simple diameter measures, which poorly represent the tumor’s true progression or regression in mesothelioma’s diffuse growth pattern. Clinicians have expressed frustration and uncertainty in evaluating treatment efficacy using these parameters, highlighting a pressing need for a more refined and reliable approach.

To address these limitations, an interdisciplinary team of AI scientists, radiologists, and pulmonologists collaborated at the Netherlands Cancer Institute. Leveraging an extensive dataset comprising over 11,000 computed tomography (CT) scans from more than 2,000 patients across 121 hospitals worldwide, they developed ARTIMES, an AI-driven volumetric response evaluation tool. Unlike humans, who face near-impossible challenges in manually delineating tumor boundaries at the pixel level on complex images, ARTIMES can effortlessly segment entire tumors and calculate their true volume with exceptional precision.

Pulmonologist Sjaak Burgers emphasizes that ARTIMES advances clinical practice by eliminating tedious and error-prone manual tumor assessments. While verifying the AI’s output remains essential, the review process is far less labor-intensive. This reduces interobserver variability and enables clinicians to obtain more consistent and objective insights into tumor dynamics. The ability to evaluate the full tumor burden rather than a single diameter mark dramatically increases sensitivity for detecting true positive or negative treatment responses.

The scientific community is witnessing a milestone with ARTIMES being the first AI model worldwide to demonstrably outperform clinicians in assessing treatment outcomes for pleural mesothelioma. Kevin Groot Lipman, lead author and technical physician, highlights that their study, published in The Lancet Oncology, cements AI’s potential to become an integral clinical decision support tool. Importantly, ARTIMES enhances rather than replaces physician judgment, interfacing smoothly into existing workflows while enabling rapid, data-driven decision-making.

Beyond measuring tumor volume, the researchers have undertaken the critical task of integrating ARTIMES measurements into actionable clinical guidelines. Since knowing the tumor size alone does not dictate specific treatment changes, these criteria empower pulmonologists to determine when to modify or cease therapies. This synergy ensures that patients receive individualized care tailored to their tumor behavior patterns, reducing exposure to ineffective treatments and unnecessary side effects while optimizing healthcare resources.

One of ARTIMES’s most transformative capabilities is its ability to detect non-response to therapy earlier than ever before. This timely recognition allows physicians to pivot treatment plans sooner, offering patients alternative therapeutic avenues or sparing them from futile and potentially harmful continuation of ineffective regimens. The combination of predictive accuracy and clinical oversight represents a major leap forward in precision oncology for pleural mesothelioma patients.

Currently, EU regulations restrict ARTIMES’s use exclusively to the Netherlands Cancer Institute under an in-house exemption, given that the model was developed internally. Nonetheless, the research team is actively pursuing regulatory approval to deploy ARTIMES globally in other hospitals. There is hopeful anticipation surrounding proposed EU frameworks aimed at streamlining the certification process for AI-enabled medical devices, which could accelerate widespread adoption and patient benefit.

The advent of ARTIMES is poised to deliver a shockwave across oncology fields by demonstrating the tangible superiority of AI over human evaluators in complex tumor assessments. To foster transparency and collaborative innovation, the Netherlands Cancer Institute has made the mesothelioma AI model publicly accessible online, enabling researchers worldwide to explore and extend its applications. This open science approach is expected to catalyze new studies and adaptations for other tumor types with challenging morphologies.

Already, the NKI team is extending their AI methodologies to address lung cancer and brain metastasis tumor evaluations. The success of ARTIMES signals the dawn of a new epoch in oncological imaging, where volumetric and morphological complexities that previously hindered precise quantification become tractable. Such breakthroughs unlock significant potential to enhance clinical trial design by furnishing robust, reproducible endpoints that more faithfully capture therapeutic impact.

Clinical trials for novel treatments stand to gain markedly from ARTIMES’s introduction. Using data from eight distinct trials, the research team validated that AI-guided volumetric criteria yield significantly improved accuracy relative to traditional RECIST assessments. This refined precision enables better evaluation of an investigational drug’s efficacy, ultimately accelerating regulatory approval timelines and facilitating faster patient access to promising therapies.

In summary, ARTIMES exemplifies how synergistic integration of AI and clinical expertise can surmount longstanding challenges in tumor measurement. By transitioning from simplistic unidimensional diameter metrics to comprehensive volumetric analysis, this technology brings unprecedented clarity and confidence to oncological decision-making. As this AI model becomes more widely disseminated and refined, it heralds a paradigm shift in cancer treatment evaluation, with rippling benefits for patients, clinicians, and research worldwide.

Subject of Research: People

Article Title: Development and validation of artificial intelligence-assisted volumetric response criteria in pleural mesothelioma (ARTIMES): a retrospective, multicohort, multicentre study

News Publication Date: 17-Jun-2026

Web References:

• The Lancet Oncology: http://dx.doi.org/10.1016/S1470-2045(26)00084-7
• Mesothelioma AI model (ARTIMES): https://huggingface.co/nki-radiology/ARTIMES
• EU Medical Devices Regulation: https://health.ec.europa.eu/medical-devices-sector/new-regulations_en

References:
Groot Lipman K, Burgers S, et al. Development and validation of artificial intelligence-assisted volumetric response criteria in pleural mesothelioma (ARTIMES): a retrospective, multicohort, multicentre study. The Lancet Oncology, 2026.

Image Credits: ©Netherlands Cancer Institute

Keywords: Cancer treatments, Artificial intelligence, Imaging analysis, Pleural mesothelioma, Tumor volumetrics, Clinical trials, Precision oncology

The study, encompassing data from over a thousand ccRCC patients collected retrospectively across five distinct medical centers, reflects one of the most comprehensive efforts to date to apply AI in this clinical context. The researchers combined data from two centers for model development and testing, while further data from three additional centers served as external validation sets, ensuring the robustness and generalizability of the models. Such a vast, multicenter dataset is crucial for training AI systems that maintain accuracy across diverse populations and imaging protocols.

Central to this endeavor were two 3D deep learning models designed explicitly for preoperative staging tasks: one targeting tumor size and extent classification (T staging) categorized into T1, T2, and combined T3 + T4 stages, and the other predicting the full TNM stage spectrum from I through IV. These models utilized corticomedullary phase CT scans, a crucial imaging phase highlighting tumor vascularity and structure, which offers rich information pivotal for staging analysis. The incorporation of a Transformer-ResNet backbone enabled the models to capture complex spatial relationships within volumetric imaging data efficiently.

Performance metrics across the validation cohorts underscored the promising capabilities of these AI tools. For T staging, the models achieved micro-average AUC values exceeding 0.93, indicating excellent overall discriminatory power, accompanied by macro-average AUCs close to 0.85 and accuracy rates surpassing 84% on average. Similarly, for TNM staging, micro-AUCs hovered around 0.93 with macro-AUCs between 0.81 and 0.89, reflecting the consistent precision across different staging categories. These results suggest that AI-assisted staging could feasibly complement or even surpass traditional radiological evaluations for preoperative staging.

However, the models showed relatively diminished accuracy when identifying advanced tumor subclasses, particularly T3 + T4 tumors and stage III in the TNM classification. AUC values in these categories ranged from approximately 0.67 to 0.80, indicating moderate performance and highlighting ongoing challenges in differentiating more complex tumor presentations. These findings emphasize the nuanced difficulties deep learning models face when confronting heterogenous and often ambiguous radiographic features typical of higher-stage tumors.

To enhance clinical interpretability, the researchers employed Gradient-weighted Class Activation Mapping (Grad-CAM), visualizing the model’s focal areas within the tumor regions during prediction. These heatmaps not only asserted that the AI anchored its decisions on clinically relevant tumor characteristics but also offered a transparent mechanism to build clinician trust in the algorithms. Interpretability remains a crucial step toward regulatory acceptance and real-world deployment of AI in healthcare.

Importantly, the study extended beyond pure algorithmic development by implementing a human-machine collaboration experiment, revealing that radiologists supported by AI predictions achieved higher diagnostic accuracy than either operating alone. This synergistic effect suggests deep learning models are not intended to replace medical expertise but rather to augment radiologists’ capabilities, reduce interobserver variability, and foster more standardized staging outcomes essential for personalized therapy.

The technical backbone of the deep learning methodology consisted of a Transformer-ResNet architecture, tailored to integrate the strengths of convolutional neural networks in feature extraction with the sophisticated contextual learning capabilities of transformer units. This hybrid design enables the model to effectively analyze 3D volumetric data, capturing the intricate texture and morphological patterns that inform tumor staging in ccRCC. Such architectural innovations represent a vital advance over previous 2D or less context-aware approaches.

While the current results indicate substantial progress, the authors acknowledge that further refinement is necessary, particularly in enhancing performance for advanced-stage tumors. This may involve incorporating multimodal imaging inputs, finer-grained subclassifications, or supplementary clinical data to enrich model context. Moreover, prospective studies and integration into clinical workflows are essential next steps to validate utility and impact in routine oncology practice.

The translational potential of this research is significant. Accurate preoperative staging guides surgical planning, eligibility for targeted therapies, and prognostication. By providing radiologists with interpretable, AI-driven insights, these models may expedite decision-making, improve patient outcomes, and reduce healthcare costs driven by diagnostic uncertainty or treatment complications. Given the global burden of kidney cancer, accessible AI tools could become invaluable in diverse healthcare settings.

This study also exemplifies the growing trend toward leveraging multi-institutional collaborations in medical AI research, which is critical to overcoming overfitting and ensuring model generalizability across populations and hardware variability. The robust external validations employed here serve as a benchmark for future investigations aiming to translate AI models from experimental settings into dependable clinical assets.

In conclusion, this seminal investigation into CT-based deep learning for ccRCC staging marks a transformative step forward in oncologic imaging. By harnessing advanced neural architectures and extensive multicenter data, the research outlines a path toward more objective, reproducible, and clinically empowering diagnostic tools. While further enhancements remain imperative, the demonstrated interpretability and human-machine synergy provide a compelling blueprint for the future integration of AI into cancer care paradigms.

As the medical community increasingly embraces artificial intelligence, studies such as these underscore the critical balance between technological innovation and clinical applicability. Future directions may explore real-time decision support during surgical and interventional procedures, integration with pathology and genomics, and expansion to other malignancies. Ultimately, AI-powered imaging analyses promise to sharpen the precision of cancer staging and personalize treatment strategies, bringing tangible benefits to patient care worldwide.

These pioneering CT-based 3D TR-Net deep learning models thus represent not only a technical achievement but also a catalyst for evolving multidisciplinary collaboration and a more nuanced understanding of cancer heterogeneity. Their deployment in clinical practice could redefine standards of care in renal oncology and potentially serve as a template for AI applications across myriad disease states.

Subject of Research: Preoperative T and TNM staging in clear cell renal cell carcinoma using CT-based deep learning

Article Title: Multicenter study of CT-based deep learning for predicting preoperative T staging and TNM staging in clear cell renal cell carcinoma

Article References:
Li, W., Xi, Y., Lu, M. et al. Multicenter study of CT-based deep learning for predicting preoperative T staging and TNM staging in clear cell renal cell carcinoma. BMC Cancer 25, 1604 (2025). https://doi.org/10.1186/s12885-025-14836-z

Image Credits: Scienmag.com

DOI: https://doi.org/10.1186/s12885-025-14836-z

The innovation pivots around a unique deep learning methodology termed the massive-training artificial neural network (MTANN). Unlike conventional AI systems, which demand vast datasets often requiring thousands or millions of annotated medical images, Suzuki’s MTANN thrives on remarkably minimal data. The system learns directly from pixel-level information extracted from computed tomography (CT) scans, drastically minimizing the training dataset to only 68 cases. This represents a colossal leap forward, circumventing the long-standing challenge of data scarcity in medical AI.

Cutting-edge deep learning architectures like Vision Transformer and 3D ResNet are usually the benchmarks in AI-based diagnosis, but Suzuki’s model outperforms these state-of-the-art (SOTA) frameworks despite their dependency on massive datasets. The MTANN approach achieved a remarkable area under the curve (AUC) of 0.92, compared to significantly lower scores from Vision Transformer and 3D ResNet models, which scored 0.53 and 0.59, respectively. This disparity underscores the efficacy of Suzuki’s approach, which also benefits from speed and portability.

Training efficiency is another standout feature of this MTANN model. Entirely trained on a commercial-grade laptop computer without specialized hardware, the process took a mere 8 minutes and 20 seconds — a fraction of the time and cost typically associated with large-scale AI training on data center infrastructures. Moreover, once trained, the system processes diagnostic predictions in just 47 milliseconds per patient case. This unprecedented speed not only streamlines clinical workflows but also expands AI’s accessibility to institutions with limited technological resources.

The implications of such technology go beyond mere cost and speed improvements. Suzuki emphasizes that this AI approach democratizes medical diagnostics, especially benefiting rare diseases where collecting extensive datasets is challenging or impossible. By reducing dependency on massive data volumes or expensive hardware, this innovation has the potential to empower healthcare providers worldwide — from well-equipped urban hospitals to rural clinics.

Another critical aspect of this breakthrough lies in its environmental impact. Conventional AI development and deployment, particularly those involving data centers filled with GPUs, pose enormous energy consumption challenges. The MTANN’s low resource demand translates into substantially reduced power usage, addressing the looming global energy concerns associated with the exponential growth in AI applications.

Suzuki’s research did not go unnoticed. At RSNA 2024, the work was honored with the prestigious Cum Laude Award, a distinction awarded to only 1.45% of all presentations. This recognition signifies the profound scientific value and impact potential embodied in the ultra-lightweight AI model — a testament to the team’s ingenuity and meticulous craftsmanship.

The MTANN concept has a storied history, dating back to Suzuki’s pioneering work in the early 2000s. Having developed one of the earliest deep learning models for medical imaging, Suzuki has continuously refined this technology over two and a half decades. With a prolific portfolio of over 400 scholarly articles and more than 40 patents — many of which have been commercialized — his work bridges the gap between cutting-edge AI theory and real-world clinical application.

His stature within the scientific community is exemplified not only by his research output but also through leadership roles, including chairing a session at the 39th Annual AAAI Conference on Artificial Intelligence. Furthermore, Suzuki received two of RSNA’s highest honors in 2024 and is ranked among the top 2% of scientists globally, cementing his influence in the AI and biomedical research arena.

Fostering interdisciplinary collaboration is a key driver behind Suzuki’s approach. Merging engineering, computer science, and medical knowledge, his team operates within a dynamic research environment that pushes the boundaries of biomedical AI. This synergy accelerates the translation of novel algorithms into practical diagnostic tools ready for real-world deployment, ensuring that theoretical breakthroughs positively impact patient care.

Looking forward, Suzuki and his collaborators envision expanding the scope of ultra-lightweight AI systems. Their work provides a blueprint for developing compact, high-performance models tailored to an array of medical imaging challenges, beyond lung cancer, and potentially other diagnostic modalities. This paradigm shift heralds a new era where AI’s power is harnessed efficiently, equitably, and sustainably.

The foundation of the Institute of Science Tokyo, officially established in October 2024 through the merger of Tokyo Medical and Dental University and Tokyo Institute of Technology, fosters this interdisciplinary and innovative atmosphere. With a mission centered on advancing science to enhance human wellbeing, the institute provides a fertile environment for breakthroughs such as Suzuki’s AI cancer diagnostic model.

In summation, the ultra-lightweight MTANN-based AI model stands as an extraordinary advancement in medical technology. It redefines the possibilities of AI diagnostics by combining efficiency, accessibility, and environmental responsibility. By enabling powerful diagnostic tools on everyday computing devices, its ripple effects could revolutionize cancer diagnosis globally, making high-quality healthcare attainable regardless of geographical or economic barriers.

Subject of Research: Artificial Intelligence for Lung Cancer Diagnosis

Article Title: Ultra-Lightweight AI Model Revolutionizes Lung Cancer Diagnosis on Standard Laptops

News Publication Date: 2024 (RSNA Annual Meeting)

Web References:

• Radiological Society of North America (RSNA) 2024 Annual Meeting: https://www.rsna.org/annual-meeting
• MTANN Deep Learning Approach: https://www.isct.ac.jp/ja/news/acnrfdt9dcto#note1
• AAAI Conference on Artificial Intelligence: https://aaai.org/conference/aaai/aaai-25/
• RSNA Highest Distinctions: https://suzukilab.first.iir.titech.ac.jp/news/news-4139/

Image Credits: Kenji Suzuki, Institute of Science Tokyo

Keywords: Cancer, Lung cancer, Artificial intelligence, Medical technology, Neural networks, Computerized axial tomography, Health and medicine, Cancer risk