B. Ashwini

B. Ashwini

Auto-generated profile, based on publicly available information

B. Ashwini

Department of Information Science and Engineering, NMAM Institute of Technology (NMAMIT), Nitte (Deemed to Be University), Karkala, Karnataka, India

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 April 2026 in IEEE Access
Verified
Journal Article
Open access
YAYeganeh Abdollahinejad
Yeganeh Abdollahinejad
SRSayed Mohsin Reza
Sayed Mohsin Reza
BAB. Ashwini
B. Ashwini
AMAhmad Mousavi
Ahmad Mousavi
Emotion recognition (ER) from physiological signals, particularly electroencephalography (EEG), is increasingly studied for healthcare uses such as mental health monitoring, stress assessment, and patient-adaptive support. Compared with behavioral observation alone, biosignals including EEG, electrocardiography (ECG), electrodermal activity (EDA), photoplethysmography (PPG; reported as blood volume pulse, BVP), and respiration provide objective correlates of affect when overt cues are limited or unreliable. This systematic review synthesizes 70 peer-reviewed studies published from 2014 to 2024. We summarize (i) physiological modalities and multimodal fusion strategies, (ii) machine learning and deep learning methods most frequently used, (iii) datasets, evaluation protocols, and metrics with explicit separation of subject-dependent (SD) and subject-independent (SI) or leave-one-subject-out (LOSO) settings, and (iv) explainable AI (XAI) practices that aim to improve interpretability and clinical trust. Across the literature, SD results are consistently higher than SI/LOSO results; domain adaptation and multimodal fusion partially close this gap. Paired within-study contrasts show small but reliable gains on DEAP for fusion over EEG alone by about 3 to 5 percentage points, and larger gains of about 25 to 35 percentage points on SEED-IV or SEED-V under participant-generic or LOSO protocols; WESAD also shows meaningful improvements. Common barriers to clinical translation include artifacts and noise, inter-subject variability, limited cohort diversity and external validation, privacy and governance constraints, and nonstandardized interpretability. We outline directions for real-world adoption, including protocol-aware benchmarking, longitudinal and cross-site validation, streamlined sensor configurations, personalization through brief calibration or domain adaptation, and clinically useful XAI that supports auditability and aligns with healthcare regulatory expectations.

30 January 2026 in IEEE Access
Verified
Journal Article
Open access
ANAnkitha A. Nayak
Ankitha A. Nayak
PVP. S. Venugopala
P. S. Venugopala
NTNicy Thomas
Nicy Thomas
MRM.S. Ravi
M.S. Ravi
BAB. Ashwini
B. Ashwini
The axial inclination of the teeth plays a significant role in orthodontics, oral surgery, and prosthodontics, ensuring the equilibrium of occlusal forces in the stomatognathic system and dental parallelism. This paper proposes a novel, automated deep learning and computer vision approach to segment and determine the mesiodistal axial inclination of individual teeth on orthopantomogram radiographs. Although the dentistry domain has explored the deep learning domain widely to solve its numerous problems, automated axial inclination and tooth morphology remain unexplored. This paper presents a customized axial inclination methodology based on deep learning and computer vision techniques that reduces time and the errors associated with manual measurements. The paper presents instance-level OPG tooth segmentation on the manually annotated dataset collected from A. B. Shetty Memorial Institute of Dental Sciences. The paper evaluates instance segmentation models using precision, recall, Dice coefficient, and IoU. The generalization of the model is justified using 5-fold cross-validation based on both object-level and pixel-level metrics. The model shows 0.83 precision, 0.92 recall, 0.87 dice coefficient, and 0.77 IoU, projecting reliable instance segmentation. Additionally, compared to existing approaches, the proposed method is reliable and faster, reducing computational overhead. The designed approach yields considerable axial inclination values, which are statistically analyzed alongside manually computed values from dentists using an unpaired t-test (p < 0.05). The study presents a clinically relevant, reliable, scalable, and structured preserving approach that provides a strong foundation for AI-based orthodontic treatment and diagnosis.

17 July 2025 in IEEE Access
Verified
Journal Article
Open access
RARamyashree
Ramyashree
PVP. S. Venugopala
P. S. Venugopala
RSRaghavendra Somappa
Raghavendra Somappa
VKVijay Kubihal
Vijay Kubihal
BAB. Ashwini
B. Ashwini
Digital watermarking in medical imaging serves as a critical technique for ensuring data integrity, authentication, and copyright protection. This study explores an innovative approach to embed diverse file types, including text, PDF, video, and audio, into DICOM images using the Least Significant Bit (LSB) and Discrete Cosine Transform (DCT) techniques. The primary objective is to evaluate the performance of these embedding techniques in terms of insertion time, extraction time, Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Mean Squared Error (MSE). To assess the robustness of the embedded data, a series of attacks are applied to the DICOM images, including zero attack, salt and pepper attack, Gaussian attack, and speckle noise attack. The impact of these attacks on the integrity and retrieval of the embedded files is meticulously analyzed. The experimental analysis reveals that under Speckle Noise Attack (SNA), the character extraction accuracy remains relatively high across all file types, highlighting its minimal impact compared to other attacks. The results demonstrate the resilience of the embedded data against various forms of attacks, providing insights into the effectiveness and security of the proposed method. Speckle Noise Attack (SNA), LSB achieves higher accuracy in text files (85.76%) and PDF files (82–85%) compared to DCT, while DCT performs better in audio files with 86–86.5% accuracy. Overall, LSB is more effective for text and PDF watermarking, whereas DCT is more robust for audio files. The proposed framework uniquely enables the embedding of diverse multimedia content into DICOM images, addressing a critical gap in existing medical watermarking studies that focus mainly on text or visual data. This multi-format capability supports hospital systems in securing comprehensive patient records such as diagnostic notes, doctor-patient consultations, and procedural videos. Furthermore, the method demonstrates strong resilience against various attacks ensuring robustness and integrity validated by superior PSNR, SSIM, and MSE metrics. This comprehensive analysis not only validates the robustness of the embedding techniques but also lays the groundwork for future research in secure medical image processing.