DeepFoci

DeepFoci automates quantification and morphometric analysis of ionizing radiation-induced foci (IRIFs) to measure DNA double-strand breaks (DSBs) and analyze DNA-repair-associated protein colocalization.

Key Features:

• Deep Learning Architecture: Utilizes a U-Net architecture for nucleus segmentation and IRIF detection complemented by maximally stable extremal region-based methods for precise IRIF segmentation.
• 3D Multichannel Data Compatibility: Trained to handle 3D multichannel datasets focusing on repair proteins 53BP1 and γH2AX.
• High Accuracy and Sensitivity: Produces IRIF quantification with accuracy comparable to the variability observed between two expert human analysts across a wide range of IRIF counts per nucleus.
• Robustness Across Diverse Conditions: Validated on challenging datasets including mixtures of nonirradiated and irradiated cells from permanent cell lines (NHDFs, U-87) and primary cultures from head and neck cancer patients, covering radiation doses of 0.5-8 Gy and post-irradiation times up to 24 hours.
• Morphometric Analysis and Colocalization: Extracts detailed three-dimensional morphometric features and assesses colocalization of repair proteins within IRIFs to enable multiparameter categorization of foci.

Scientific Applications:

• Radiation Biodosimetry: Provides precise IRIF quantification to support assessment of radiation exposure.
• Research on DNA Damage and Repair: Facilitates analysis of DSB induction and repair processes at molecular and single-cell levels.
• Radiotherapy Monitoring: Enables monitoring of IRIF dynamics relevant to radiotherapy treatment evaluation.
• Tumor Classification and Subclone Identification: Supplies multiparameter IRIF-derived features to aid tumor classification and identification of cell subclones.

Methodology:

Implements a U-Net for nucleus segmentation and IRIF detection, uses maximally stable extremal region-based methods for IRIF segmentation, is trained on 3D multichannel data for 53BP1 and γH2AX, and performs three-dimensional morphometric feature extraction and colocalization analysis.