سامانه همایش ها

English

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Integration of High-Speed AFM Nanomechanical Profiling with Deep Spatiotemporal Learning for Early Response Assessment and Tumor Stage Prediction in Oncolytic Virotherapy

نویسندگان :

َAlireza Haghighatjoo¹ Fatemeh Noori² Peyman Afshari Bijarbaneh³ Seyed Amirhossein Mousavi⁴

1- دانشگاه آزاد اسلامی واحد مشهد 2- Genova University , italy 3- Genova University , italy 4- دانشگاه ازاد مشهد

کلمات کلیدی :

Oncolytic Virotherapy،Atomic Force Microscopy،Early Response Biomarkers،Deep Learning / CNN-LSTM،Nanomechanical Profiling

چکیده :

Oncolytic virotherapy has emerged as a promising therapeutic strategy in oncology, utilizing genetically engineered or naturally occurring viruses to selectively infect, replicate within, and lyse malignant cells while sparing normal tissues. Beyond direct cytolysis, oncolytic viruses stimulate antitumor immunity by releasing tumor-associated antigens and danger signals, thereby initiating systemic immune responses. However, clinical translation is limited by the lack of robust early-stage biomarkers for treatment response assessment and patient stratification. Conventional evaluation methods—imaging, molecular assays, and immunohistochemistry—often provide delayed or indirect feedback, restricting timely therapeutic optimization. Nanomechanical profiling of living cells offers a novel approach for real-time assessment of therapeutic efficacy. Atomic Force Microscopy (AFM) enables label-free, nanoscale measurement of cellular mechanical properties, including stiffness, elasticity, and surface topography, which dynamically change in response to cytoskeletal remodeling and therapeutic interventions. Recent advances in high-speed AFM, coupled with artificial intelligence (AI), allow automated, high-dimensional feature extraction from raw nanomechanical data. Deep learning frameworks, particularly hybrid convolutional neural network (CNN) and long short-term memory (LSTM) architectures, have been shown to accurately classify cancer-related phenotypes and detect extracellular vesicle signatures with AUC values exceeding 0.90. In this study, we present a novel framework combining high-speed AFM with a hybrid CNN–LSTM pipeline to assess early responses to oncolytic virotherapy. Using human tumor cell lines infected with clinically relevant oncolytic viruses (HSV-1, VSV, NDV), we demonstrate that nanomechanical signatures acquired within the first 24 hours reliably distinguish responders from non-responders with >95% accuracy and predict tumor stage with AUC values of 0.90–0.95. Responding cells exhibited early softening, increased nanovibrational activity, and membrane roughening, whereas non-responders displayed attenuated mechanical changes, providing distinct mechanobiological fingerprints of therapeutic efficacy.Our findings establish AFM–AI integration as a rapid, label-free, and non-destructive platform for early response monitoring and tumor stage prediction in oncolytic virotherapy. This approach surpasses conventional methods in speed, sensitivity, and specificity, offering a translational pathway toward real-time patient stratification and personalized treatment adjustments. The study lays the groundwork for further integration of mechanobiology-informed biomarkers with complementary omics and imaging modalities to advance precision oncology workflows.