Purpose: To propose an automated approach for detecting and classifying Intracranial Hemorrhages (ICH) directly from sinograms using a deep learning framework. This method is proposed to overcome the limitations of the conventional diagnosis by eliminating the time-consuming reconstruction step and minimizing the potential noise and artifacts that can occur during the Computed Tomography (CT) reconstruction process.

 Methods: This study proposes a two-stage automated approach for detecting and classifying ICH from sinograms using a deep learning framework. The first stage of the framework is Intensity Transformed Sinogram Synthesizer, which synthesizes sinograms that are equivalent to the intensity-transformed CT images. The second stage comprises of a cascaded CNN-RNN model that detects and classifies hemorrhages from the synthesized sinograms. The CNN module extracts high-level features from each input sinogram, while the RNN module provides spatial correlation of the neighborhood regions in the sinograms. The proposed method was evaluated on a publicly available RSNA dataset consisting of a large sample size of 8,652 patients. 

Results: The results showed that the proposed method had a notable improvement of as high as 27% in patient-wise accuracies when compared to state-of-the-art methods like ResNext-101, Inception-v3 and Vision Transformer. Furthermore, the sinogram-based approach was found to be more resilient to noise and offset errors in comparison to CT image-based approaches. The proposed model was also subjected to a multi-label classification analysis to determine the hemorrhage type from a given sinogram. The learning patterns of the proposed model were also examined for explainability using the activation maps.

Conclusion: The proposed sinogram-based approach can provide an accurate and efficient diagnosis of ICH without the need for the time-consuming reconstruction step and can potentially overcome the limitations of CT image-based approaches. The results show promising outcomes for the use of sinogram-based approaches in detecting hemorrhages, and further research can explore the potential of this approach in clinical settings.

In recent years, the medical diagnostic field has shown significant interest in deep learning due to its ability to identify challenging patterns accurately. While conventional computer-aided diagnosis (CAD) systems primarily relied on human-interpretable CT scans, the direct utilization of sinograms for diagnosis is not explored. This paper introduces a novel approach to hemorrhage detection in the brain by combining sinograms, the raw projections that are acquired before CT reconstruction, with the reconstructed CT images.  The proposed model, "SinoCTFusionNet," was evaluated for intracranial hemorrhage detection using the publicly available RSNA dataset. The integration of both CT images and sinograms yielded notable advancements. Specifically, the combined approach achieved a 3.38\% improvement in detection accuracy compared to CT-based detection, as well as a substantial 16.82\% enhancement compared to sinogram-based detection. Further, the model's robustness to noise and offset errors incurred during CT image reconstruction is investigated. It is observed that the proposed fusion model exhibited robustness against such introduced errors, while the performance of the single-input model experienced a decline.