A deep learning method using SPECT images to diagnose remaining thyroid tissue post-thyroidectomy

RR‐HCL‐SVM: A two‐stage framework for assessing remaining thyroid tissue post‐thyroidectomy in SPECT images

“…The average F1 score achieved is 0.97 on the same dataset as in 6 and. 1 These features are also aligned with the diagnostic process of physicians, as in reality, the presence of surplus bone armor in patients depends on the radiation correlation and the absorptive characteristics of the cervical region.…”

“…This is demonstrated in the paper "Deep Learning Method Using SPECT Images to Diagnose Remaining Thyroid Tissue Post-Thyroidectomy". 1 Following these experiments, when working with a larger dataset and a more substantial model, notable enhancements in classification performance were observed. This substantiates the viability of incorporating AI into the diagnostic process, suggesting that the application of a sufficiently expansive model along with an extensive dataset could potentially facilitate the classification of residual and non-residual thyroid tissue solely based on SPECT images.…”

“…This spatial awareness has made DETR a groundbreaking advancement in the field of computer vision, offering a more streamlined and effective approach to object detection. The study conducted by the previous study 7 demonstrates that the DETR model yields outstanding results compared to other CNN-based detectors such as Faster-RCNN and YoloV7. In their research, the DETR model achieved a remarkable mAP@0.5 score of 0.43 on a dataset with characteristics similar to the dataset used in this study.…”

“…This substantiates the viability of incorporating AI into the diagnostic process, suggesting that the application of a sufficiently expansive model along with an extensive dataset could potentially facilitate the classification of residual and non-residual thyroid tissue solely based on SPECT images. 1 With these promising results, the deep learning method for detecting the location of residual thyroid tissue post-surgery has been implemented, extending beyond mere classification into residual or non-residual categories. 7 Based on existing research results, we want to improve and develop an end-to-end flow to support doctors with high accuracy in patient screening.…”

“…1 With these promising results, the deep learning method for detecting the location of residual thyroid tissue post-surgery has been implemented, extending beyond mere classification into residual or non-residual categories. 7 Based on existing research results, we want to improve and develop an end-to-end flow to support doctors with high accuracy in patient screening.…”