Akademische Forschung
Unsere Arbeit basiert auf peer-reviewed Forschung. Nachfolgend ausgewählte Publikationen aus der akademischen Karriere von Dr. Dario Sitnik in medizinischer Bildgebung, Computer Vision und KI.
// Medical Imaging
Deep Learning-Based Mitosis Detection in Breast Cancer Histopathology Images
D. Sitnik, M. Aubreville, A. Maier
International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2023
We present a novel deep learning approach for automated mitosis detection in H&E-stained breast cancer histopathology slides, achieving state-of-the-art performance on the MIDOG challenge dataset.
Multi-Scale Feature Aggregation for Histopathological Image Classification
D. Sitnik, C. Bertram, R. Klopfleisch, A. Maier
IEEE Transactions on Medical Imaging, 2023
A multi-scale feature aggregation framework for whole-slide image classification that captures both cellular-level and tissue-level morphological patterns in histopathology.
Domain Adaptation for Cross-Scanner Histopathology Image Analysis
D. Sitnik, M. Aubreville, A. Maier
Medical Image Analysis (MedIA), 2022
We address the domain shift problem in computational pathology by proposing a domain adaptation method that generalizes across different scanners and staining protocols.
// Computer Vision
Attention-Guided Object Detection for Industrial Quality Control
D. Sitnik, F. Wagner, A. Maier
European Conference on Computer Vision (ECCV) Workshops, 2022
An attention-guided detection framework optimized for manufacturing defect detection, achieving real-time performance on edge devices while maintaining high accuracy on small defects.
Few-Shot Learning for Visual Inspection in Manufacturing
D. Sitnik, A. Maier
Pattern Recognition (Elsevier), 2022
A few-shot learning approach for visual defect detection that requires only 5-10 labeled examples per defect class, making AI-based quality control practical for small-batch manufacturing.
// Document Intelligence
Layout-Aware Transformer for Structured Document Information Extraction
D. Sitnik, J. Pfeiffer, A. Maier
International Conference on Document Analysis and Recognition (ICDAR), 2023
A layout-aware transformer architecture for extracting structured information from complex documents such as contracts, invoices, and medical records with state-of-the-art accuracy.
Privacy-Preserving Document Processing with Federated Learning
D. Sitnik, A. Maier
ACL Workshop on Privacy in NLP, 2023
A federated learning framework for document processing that enables model training across organizations without sharing sensitive document data, achieving comparable accuracy to centralized approaches.
// Natural Language Processing
Multilingual Legal Document Analysis with Cross-Lingual Transfer
D. Sitnik, A. Maier
Findings of EMNLP, 2023
Cross-lingual transfer learning for legal document analysis across German, English, and Croatian, demonstrating effective knowledge transfer for contract clause classification and risk assessment.
Bereit loszulegen?
Buchen Sie eine kostenlose Beratung, um Ihr KI-Projekt zu besprechen.