Multimodal Approach Research Articles

Combining functional, structural, and morphological networks for multimodal classification of developing autistic brains.

Accumulating neuroimaging evidence suggests that abnormal functional and structural brain connectivity plays a cardinal role in the pathophysiology of autism spectrum disorder (ASD). Here, we constructed brain networks of functional, structural, and morphological connectivity using data from functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and structural magnetic resonance imaging (sMRI), respectively. The neuroimaging data from a cohort of 50 individuals with ASD and 47 age-, gender- and handedness-matched TDC (age range: 5-18 years) were selected from the Autism Brain Image Data Exchange database. The combination of the fMRI, sMRI and DTI modalities connectivity features resulted in a classification accuracy of 82.69% for differentiating individuals with ASD from TDC. This accuracy surpassed that of any single modality or combination of fMRI and DTI modalities previously examined. Among the fMRI, sMRI and DTI modalities, the most distinguishing connectivity features were observed in the temporal, parietal, and occipital lobes from the DTI modality, the prefrontal and parietal lobes from the fMRI modality, and the temporal lobe from the sMRI modality. In addition, we also found that these distinguishing connectivity features can predict abnormal social interaction behaviours in ASD. These results highlight the complementary information provided by multimodal approaches, further emphasizing the pivotal role of multimodal connectivity patterns in unravelling the intricate mechanisms involved in the pathophysiology of ASD.

On-shift resident education about hematologic and oncologic emergencies: A needs-based assessment.

9020 Background: Hematology/oncology (H/O) is a fundamental topic for the American Board of Internal Medicine (ABIM); however, there are few resources designed for resident-level education. While asynchronous approaches have gained popularity, there can be a synergistic effect to cementing understanding during clinical exposure. Rotations provide this opportunity, but limited tools exist to augment practice-based learning. Multimodal approaches, such as self-directed text and near-peer educational models, are rarely accessible on shift. The breadth and depth of such resources can make it difficult for users to distill salient information during patient care. This needs-based assessment (NBA) sought to characterize residents’ attitudes towards on-shift learning as well as their comfort with diagnosing and treating 3 common H/O emergencies. Methods: An NBA survey was sent to 104 internal medicine (IM) and medicine-pediatrics (MP) residents at an academic medical center. Descriptive statistics are reported. Results: Fifty-five of 104 residents completed the survey, for a response rate of 53%. Forty-seven percent were post-graduate year (PGY) 1, 25.5% were PGY-2, 25.5% were PGY-3 and 2% were PGY-4. While 98% of respondents use educational resources on shift, only 6% had a H/O specific resource. The most used resources were Up-to-Date (98%), the Mass General Hospital WhiteBook (75%), PubMed (47%), and artificial intelligence (AI) (47%). The biggest barriers to resource utilization were lack of time due to clinical responsibilities (76%) and length of resource (69%). Ninety-one percent indicated interest in a resource designed for shift-based learning and would most value guideline inclusion (80%). While most residents (62%) were comfortable diagnosing tumor lysis syndrome (TLS) and febrile neutropenia (FN) (69%), they were neutral/uncomfortable with diagnosing hyperleukocytosis (71%). Most were comfortable treating FN (53%) but were neutral/uncomfortable with treating TLS (60%) or hyperleukocytosis (96%). There was a statistically significant difference between PGY-1 and PGY-2+ in comfort with diagnosis (p=.0018) and treatment (p<.001) of FN as well as the treatment (p=.0132) of TLS. There was no significant difference in comfort with TLS diagnosis, or hyperleukocytosis diagnosis or treatment among PGY. Conclusions: This assessment demonstrates an overwhelming interest in an easily accessible, guidelines-based, electronic resource for residents to utilize during clinical care. Though a medical emergency, hyperleukocytosis is a H/O diagnosis that our program’s residents are not comfortable identifying or managing. These results highlight the opportunity for growth in H/O education of IM/MP residents and will guide the design of a digital education resource, as well as its implementation and evaluation at our institution.

Evaluating AI-generated patient education materials for spinal surgeries: Comparative analysis of readability and DISCERN quality across ChatGPT and deepseek models.

Access to patient-centered health information is essential for informed decision-making. However, online medical resources vary in quality and often fail to accommodate differing degrees of health literacy. This issue is particularly evident in surgical contexts, where complex terminology obstructs patient comprehension. With the increasing reliance on AI models for supplementary medical information, the reliability and readability of AI-generated content require thorough evaluation. This study aimed to evaluate four natural language processing models-ChatGPT-4o, ChatGPT-o3 mini, DeepSeek-V3, and DeepSeek-R1-in generating patient education materials for three common spinal surgeries: lumbar discectomy, spinal fusion, and decompressive laminectomy. Information quality was evaluated using the DISCERN score, and readability was assessed through Flesch-Kincaid indices. DeepSeek-R1 produced the most readable responses, with Flesch-Kincaid Grade Level (FKGL) scores ranging from 7.2 to 9.0, succeeded by ChatGPT-4o. In contrast, ChatGPT-o3 exhibited the lowest readability (FKGL>10.4). The DISCERN scores for all AI models were below 60, classifying the information quality as "fair," primarily due to insufficient cited references. All models achieved merely a "fair" quality rating, underscoring the necessity for improvements in citation practices, and personalization. Nonetheless, DeepSeek-R1 and ChatGPT-4o generated more readable surgical information than ChatGPT-o3. Given that enhanced readability can improve patient engagement, reduce anxiety, and contribute to better surgical outcomes, these two models should be prioritized for assisting patients in the clinical. This study is limited by the rapid evolution of AI models, its exclusive focus on spinal surgery education, and the absence of real-world patient feedback, which may affect the generalizability and long-term applicability of the findings. Future research ought to explore interactive, multimodal approaches and incorporate patient feedback to ensure that AI-generated health information is accurate, accessible, and facilitates informed healthcare decisions.

Non-invasive diagnosis of lung diseases via multimodal feature extraction from breathing audio and chest dynamics.

Non-invasive diagnosis of lung diseases via multimodal feature extraction from breathing audio and chest dynamics.

Advances in DeepFake detection algorithms: Exploring fusion techniques in single and multi-modal approach

Advances in DeepFake detection algorithms: Exploring fusion techniques in single and multi-modal approach

Phase II study of pembrolizumab in combination with cisplatin or carboplatin and pemetrexed as induction chemoimmunotherapy in resectable epithelioid and biphasic pleural mesothelioma (CHIMERA study).

TPS8122 Background: Pleural mesothelioma (PM) is a rare cancer related to asbestos exposure, marked by complex histopathological diagnosis and dismal prognosis. Patients’ survival is strongly influenced by the histological subtype and by the eligibility to a multimodal approach, which is reserved to very selected patients. Platinum-pemetrexed chemo-regimen or the immunotherapy combination ipilimumab+nivolumab are the available first-line treatment options for unresectable PM patients. In this setting, pembrolizumab in combination with platinum-pemetrexed showed an improved overall and progression free survival (IND227/Keynote483 trial). In patients with resectable PM, the multimodality approach with platinum-pemetrexed chemotherapy and surgery is usually preferred, achieving pathological complete response (pCR) in 5% of cases. To date, the role of perioperative immunotherapy for PM has not yet been extensively investigated. Methods: This is a phase II single arm trial enrolling patients with resectable PM from 8 high volume Italian centers, with 18 months of enrollment and 12 months of follow-up. Inclusion criteria will be the histologically confirmed diagnosis of surgical resectable stage I-IIIA treatment-naïve epithelioid/biphasic PM. Patients will receive 3 cycles of pembrolizumab 200 mg plus cisplatin (75 mg/sm) or carboplatin (AUC 5) and pemetrexed (500 mg/sm) every 3 weeks. The surgical procedure of pleurectomy/decortication will be centralized in 2 centers and will be performed within 6 weeks after the last neoadjuvant cycle. The adjuvant treatment will start within 10 weeks from surgery and will be based on 14 cycles of pembrolizumab 200 mg every 3 weeks. The primary endpoint will be the pCR; secondary endpoints will include: major pathological response, objective response rate, event free survival, OS, surgery feasibility, safety. Translational analysis on tissue and blood samples will also be performed. In order to investigate an improvement of pCR from 5% to 18%, 36 patients and a minimum number of 4 pCR are needed to verify this hypothesis with a least 80% power and a probability of type I error of 0.05. Considering a 10% patients dropped-out because of disease progression precluding surgery, a total number of 40 patients will be included in the study. The trial is currently ongoing since November 2024; 5 patients have been enrolled so far. This is the first clinical trial assessing the activity and safety of pembrolizumab in combination with platinum-pemetrexed for resectable PM patients. Clinical trial information: NCT06155279 .

Effect of intratumor dendritic cell vaccination with and without chemoradiation in induced oral squamous cell carcinoma of hamsters.

Effect of intratumor dendritic cell vaccination with and without chemoradiation in induced oral squamous cell carcinoma of hamsters.

Automated Targeted Sectioning of Resin-embedded Hard Tissue Specimen Using Micro-computed Tomography in Combination with Laser Microtomy.

Histological analysis of hard tissue specimens is widely used in clinical practice and preclinical research, but it remains a labor-intensive and destructive process. In particular, resin-embedded tissues present challenges due to the inability to target regions of interest (ROI), as internal structures are not visible externally. This work proposes a guided sectioning workflow that enables precise targeting of concealed ROIs using a multimodal approach. By combining microCT imaging with an automated cutting system, and laser microtomy, precise targeted sectioning was achieved. MicroCT imaging enables visualization of internal structures, guiding the automated cutting system for precise sectioning. Laser microtomy then allows thin tissue sections to be prepared while preserving diagnostic features. Comparing the automated workflow to the conventional cutting-grinding technique showed that the new method improved accuracy by a factor of 7 and reduced material loss by half and processing time by 75%. Validation was performed by comparing the histological sections with in silico target planes generated from the microCT scans, showing precise alignment between the targeted regions and the prepared sections. We demonstrate that the proposed approach significantly reduces tissue loss and offers a more efficient workflow compared to traditional methods. Additionally, microCT-based targeting enables accurate correlation between histological findings and 3D pathological structures. The automated guided sectioning workflow provides valuable insights into tissue pathology, enhancing clinical diagnostics and preclinical research. It also facilitates the generation of multimodal datasets, which can be used in future machine learning applications.

Artificial intelligence in fetal brain imaging: Advancements, challenges, and multimodal approaches for biometric and structural analysis.

Artificial intelligence in fetal brain imaging: Advancements, challenges, and multimodal approaches for biometric and structural analysis.

Treating head and neck venous malformations with cold helium plasma electrosurgical device: A 17 patients case series.

Treating head and neck venous malformations with cold helium plasma electrosurgical device: A 17 patients case series.

StopSpamX: A multi modal fusion approach for spam detection in social networking.

StopSpamX: A multi modal fusion approach for spam detection in social networking.

A multimodal hierarchical learning approach for virtual metrology in semiconductor manufacturing

A multimodal hierarchical learning approach for virtual metrology in semiconductor manufacturing

A multimodal mixed-methods approach for holistic insights in Roux-en-Y gastric bypass patients: Protocol for a patient-centered framework

A multimodal mixed-methods approach for holistic insights in Roux-en-Y gastric bypass patients: Protocol for a patient-centered framework

Identifying offensive memes in low-resource languages: A multi-modal multi-task approach using valence and arousal

Identifying offensive memes in low-resource languages: A multi-modal multi-task approach using valence and arousal

Effect of obesity on the survival outcomes among Indian women with breast cancer.

e12558 Background: Studies have reported increased risk of breast cancer (BC) with obesity in Indian population and few studies also highlighted the association of Body Mass Index (BMI) with BC types and chemotherapy response. However, no study till date has reported the effect of obesity on the survival outcomes among Indian women with BC. Methods: We retrospectively analyzed the medical records of women registered at our institute with a diagnosis of BC between years 2013-2024. Along with demographic, clinical and tumor characteristics, baseline weight and height were retrieved from the medical records. Primary outcomes were disease-free survival (DFS) and overall survivals (OS). Cox-regression and Kaplan-Meier survival analysis were used to examine the association of BMI with survival outcomes. Results: Study sample consisted of 1303 women with BC. Median age of the study cohort is 46years and 42% were postmenopausal. Fifty-one percent of the cases were Hormone receptor (HR) positive and Her2Neu negative, followed by 29.4% (HR+ and Her2Neu+), 14.5% (TNBC) and 5.3% (HR- and Her2Neu+). AJCC-Stage I & II tumors consisted of 42.6%, followed by 41.1% Stage-III and 16.3% Stage IV tumors. Forty-one percent were obese (BMI>27.5kg/m 2 ) and 35% were overweight (BMI: 23-27.5kg/m 2 ). Fifty percent of postmenopausal women were obese at time of presentation compared to 34.5% of premenopausal women (p<0.001). Proportions of those with ECOG score ≥2 and with one or more comorbidity were significantly high among obese group compared to other BMI categories (p<0.05). Family history of cancer, clinical stage, HR/Her2Neu receptor status and type of surgery received did not vary across the BMI categories. Median DFS for overweight (110 months) and obese (107.2 months) individuals was significantly lower compared to normal/underweight individuals (194 months). After adjusting for menopausal status, comorbidity, clinical stage and receptor status; BMI≥27.5 was significantly associated with poor DFS [aHR(95%CI):1.78(1.25-2.55), p=0.002]. Premenopausal status, Stage>II and ECOG score≥2 were associated with poor DFS. Among those with non-metastatic breast cancer at time of presentation, BMI≥27.5 was associated with poor overall survival [aHR(95%CI):1.91(1.01-3.63), p=0.047], however, BMI was not associated with overall survival among those who with metastatic breast cancer. The effect of BMI on the survival outcomes didn’t vary by menopausal status and receptor status. Conclusions: BMI is independent predictor of survival among Indian women with BC. Clinical trials with multimodal therapeutic approaches including nutritional and lifestyle interventions are required to improve the survival outcomes among obese women with BC.

A multimodal embedding transfer approach for consistent and selective learning processes in cross-modal retrieval

A multimodal embedding transfer approach for consistent and selective learning processes in cross-modal retrieval

Optimizing EEG - based Emotion Recognition with a multi-modal ensemble approach

Optimizing EEG - based Emotion Recognition with a multi-modal ensemble approach

Copper-based hollow mesoporous nanogenerator with reactive oxygen species and reactive nitrogen species storm generation for self-augmented immunogenic cell death-mediated triple-negative breast cancer immunotherapy.

Copper-based hollow mesoporous nanogenerator with reactive oxygen species and reactive nitrogen species storm generation for self-augmented immunogenic cell death-mediated triple-negative breast cancer immunotherapy.

Relevance and specificity of the French CIAN LUNG cohort for clinical research analyses in lung cancer using real-world data.

e23314 Background: Lung cancer, particularly non-small cell lung cancer (NSCLC), remains a leading cause of cancer-related mortality worldwide. Conducting robust clinical research is often hindered by lengthy study timelines and fragmented data. The Cancer Imaging Analytics Network (CIAN) LUNG cohort is a unique French real-world data (RWD) initiative that integrates data from multiple centers to accelerate research. This cohort enables comprehensive longitudinal analyses, facilitating the generation of real-world evidence (RWE) reflective of routine clinical practice. CIAN LUNG includes multimodal data comprising clinical, imaging, and biomarker information. This allows descriptive analyses based on lung cancer subtypes, tumor characteristics, treatment pathways, and patient outcomes. With a multicentric methodology and long-term follow-up, the cohort aims to support predictive modeling, decision-making processes, and the validation of imaging biomarkers, ultimately contributing to precision oncology. Methods: Patients diagnosed with SCLC or NSCLC since January 2015 are eligible, with inclusion criteria requiring pathologically confirmed diagnosis, at least one CT scan at baseline, treatment at participating centers, age ≥18 years, and informed consent. Exclusion criteria include lung cancers of non-epithelial origin such as lung sarcoma or lymphoma. The inclusion of patient data to the cohort adheres to ethical guidelines, including the European Convention on Human Rights and the General Data Protection Regulation. Data extraction from patient medical records to the CIAN cohort is conducted via a hybrid approach: 1- automated processes for structured data (e.g., chemotherapy, lab results); 2-manual processes for unstructured data, curated by experienced clinical research associates. Imaging data undergo de-identification through irreversible modification of DICOM metadata, to ensure compliance with privacy regulations. Upon patient inclusion, a minimum variable product data catalog is completed and updated for sub-projects. Results: As of January 2025, approximately 1,250 patients have been enrolled across three active centers (two cancer centers and one hospital). The cohort is currently operational for: rapid screening and sub-cohort selection, data extraction across multiple sites for research purposes; serving as external control arms in clinical trials and supporting scientific publications and presentations. Conclusions: The CIAN LUNG cohort provides a robust and comprehensive RWD resource for lung cancer research. Its unique multicentric, multimodal, and longitudinal approach enhances research efficiency, offering valuable insights into tumor characteristics, treatment pathways, and patient outcomes.

Enhancing computational thinking assessment: A multimodal cognitive diagnostic approach

Enhancing computational thinking assessment: A multimodal cognitive diagnostic approach