Kidney stone detection in CT scans using attention based multi-path feature fusion networks

Detecting kidney stones in CT images presents significant challenges due to variations in stone size,
shape, intensity, and their similarity to surrounding tissues. Traditional methods often struggle with
false positives and missed detections, especially in complex or noisy scan environments. To address
these issues, we propose a novel deep learning architecture that combines advanced feature
extraction, attention mechanisms, and multi-scale fusion strategies. The model incorporates the
REPNCSPELAN4 block to capture diverse spatial and channel-wise features, followed by an ADown
module for aggressive down sampling, enabling deeper semantic understanding with efficient
computation. The SPEELAN block introduces spatial and channel attention to highlight
diagnostically relevant regions,while the CBFuse module performs cross-block fusion, integrating
fine-grained details with high-level context for improved multi-scale detection. Experimental
evaluations demonstrate that the proposed model achieves a precision of 0.798, recall of 0.742, and
mAP of 0.795, showing its effectiveness and robustness in accurately detecting kidney stones across
diverse CT scenarios.