Agenda

MSc SPS Thesis presentation

• Monday, 10 March 2025
• 14:00
• HB17.140 (Seminar room)

Surgical Workflow Analysis, an Explainable Approach

Surgical workflow analysis is crucial in optimizing procedural efficiency, resource utilization, and patient safety in catheterization laboratories. Traditional manual workflow analysis methods are labour-intensive and prone to inconsistencies, prompting the need for automated solutions that leverage machine learning and computer vision. This thesis presents an explainable two-stage model for surgical workflow analysis using ceiling-mounted cameras.

The proposed approach integrates a YOLOv8 object detection model with a Gaussian Mixture Model - Hidden Markov Model (GMM-HMM) framework. The first stage detects key objects that are input to the second stage. The GMM-HMM component then infers surgical workflow phases by modelling spatial and temporal dynamics, enabling real-time phase classification. The model is validated on two datasets from different hospitals, achieving a classification accuracy of 95.2% for the RDGG dataset and 95.4% for the Tampere dataset, ensuring its generalisability across diverse clinical environments. Experimental results demonstrate that the model is highly accurate in detecting workflow phases, emphasizing explainability and robustness. Combining YOLOv8’s efficient object detection with GMM-HMM’s structured temporal inference ensures that phase transitions are identified with minimal error. The model’s real-time feasibility and generalization across hospitals highlight its potential for clinical implementation.

This research advances automated surgical workflow analysis by addressing the dual challenges of interpretability and adaptability. Future work includes enhancing the model’s robustness to occlusions, integrating additional modalities such as audio data, and exploring its application in other surgical environments.

Seminar Tadashi Ebihara (University of Tsukuba, Japan)

• Monday, 10 March 2025
• 15:30-16:30
• EEMCS, HB. 17.140

Underwater Acoustic Communication and Positioning in Shallow Waters

Underwater Acoustic Communication and Positioning in Shallow Waters

This study addresses the challenges of underwater acoustic (UWA) communication and positioning in shallow waters, which are characterized by significant delay and Doppler spreads. To provide a highly reliable communication, we propose Doppler-resilient orthogonal signal-division multiplexing (D-OSDM), which preserves orthogonality among data vectors even in doubly spread channels. Field tests, including MIMO configurations, demonstrate that D-OSDM enables reliable data transfer in mobile UWA environments with delay and Doppler spreads. Additionally, a new underwater positioning method is proposed for shallow waters, using direct wave arrival time groups and database matching. This method accurately measures baseline length from the impulse response of the UWA channel. Experimental trials in a harbor area showed positioning errors of about 0.2 meters and a near 0% missing rate, indicating the potential of this system for unmanned and remote underwater construction.

EURASIP Webinar

• Wednesday, 12 March 2025
• 15:00-16:00
• (zoom link)

AI for applications in psychiatry

Abstract:
In this talk, we will consider applications of AI in the domain of psychiatry. Specifically, we will give an overview of our research towards automated behavioral analysis for assessing psychiatric symptoms. Schizophrenia (SCZ) and depression (MDD) are two chronic mental disorders that seriously affect the quality of life of millions of people worldwide. We aim to develop machine-learning methods with objective linguistic, speech, facial, and motor behavioral cues to reliably predict the severity of psychopathology or cognitive function, and distinguish diagnosis groups. We collected and analyzed the speech, facial expressions, and body movement recordings of 228 participants (103 SCZ, 50 MDD, and 75 healthy controls) from two separate studies. We created an ensemble machine-learning pipeline and achieved a balanced accuracy of 75.3% for classifying the total score of negative symptoms, 75.6% for the composite score of cognitive deficits, and 73.6% for the total score of general psychiatric symptoms in the mixed sample containing all three diagnostic groups. The proposed system is also able to differentiate between MDD and SCZ with a balanced accuracy of 84.7% and differentiate patients with SCZ or MDD from healthy controls with a balanced accuracy of 82.3%. These results suggest that machine-learning models leveraging audio-visual characteristics can help diagnose, assess, and monitor patients with schizophrenia and depression.

Speaker:
Dr. Justin Dauwels is an Associate Professor at the TU Delft (Signals and Systems, Department of Microelectronics), and serves as co-Director of the Safety and Security Institute at the TU Delft. He also is the scientific lead of the Model-Driven Decisions Lab (MoDDL), a first lab for the Knowledge Building program between the Netherlands police and the TU Delft.
His research interests are in data analytics with applications to intelligent transportation systems, autonomous systems, and analysis of human behavior and physiology. His academic lab has spawned four startups across a range of industries, ranging from AI for healthcare to autonomous vehicles.

Link to the webinar: https://us02web.zoom.us/j/87653221538?pwd=g8J1c2iwOKMrFZNsQT2d2PZ7R9W8HW.1

ME colloquium

• Thursday, 20 March 2025
• 16:00-17:00
• EEMCS, lecture hall Chip

Bridging Communications, Sensing and Graphs in Signal Processing

Prof. Leus will provide an overview of key research topics within the Signal Processing Systems group, spanning communications, sensing, and graph signal processing. In the field of communications, The work on a specific modulation format for underwater acoustic communications will be discussed. This type of signal modulation has been independently developed by different research groups under various names, including V-OFDM, OSDM, and OTFS. Regarding sensing, He will present our coded cover designs to boost the performance of acoustic vector sensors, which can also be applied to other types of sensors and sensor arrays. Additionally, he will share our latest findings on optimal array and waveform design for co-located multiple-input multiple-output (MIMO) radar. Interestingly, these designs deviate from conventional notions of optimality. Finally, prof. Leus will introduce the field of graph signal processing, highlight our pioneering contributions, and outline potential directions for future research.