<b><i>Background:</i></b> Axillary staging in patients with complete response after neoadjuvant chemotherapy (NAC) is still controversial. Our objective was to test tattoo alone and subsequentially tattoo plus clip as markers in the targeted axillary dissection of ycN0 patients. <b><i>Methods:</i></b> Prospective cohort of cT1-T3, cN1 (proven histologically), M0 patients scheduled to receive NAC. Exclusion criteria were lobular histology, prior axillary surgery, and clinical N2/3. In cohort 1 this positive node (Neotarget node) was tattooed at diagnosis. If ycN0, a targeted axillary dissection was performed. After an interim analysis with negative results we changed the protocol in order to do a double marking procedure (Cohort 2): the positive node was clipped at diagnosis and after NAC a tattoo was done before surgery. <b><i>Results:</i></b> Thirteen patients in Cohort 1 and 18 patients in Cohort 2. Failure to identify the Neotarget node with multiple nodes retrieved in 9/13 (69%) of Cohort 1 patients. Also in 5/13 (38%) of Cohort 1 patients and 3/18 (17%) of Cohort 2 there was a failure to clearly identify tattooed nodes. In Cohort 2, clip identification by surgical specimen radiography allowed the identification of the tagged node in 17/18 (94,4%) of cases. The concordance between the clipped node and sentinel nodes was 16/18 (89%). <b><i>Conclusions:</i></b> The introduction of double marking by clipping the metastatic node and verifying their removal by surgical specimen radiography, using carbon ink as a tracer, allowed the identification of the metastatic node in 94% of cases, with a simple, reproducible, and easy-to-implement targeted axillary dissection procedure.

The scarcity of balanced and annotated datasets has been a recurring problem in medical image analysis. Several researchers have tried to fill this gap employing dataset synthesis with adversarial networks (GANs). Breast magnetic resonance imaging (MRI) provides complex, texture-rich medical images, with the same annotation shortage issues, for which, to the best of our knowledge, no previous work tried synthesizing data. Within this context, our work addresses the problem of synthesizing breast MRI images from corresponding annotations and evaluate the impact of this data augmentation strategy on a semantic segmentation task. We explored variations of image-to-image translation using conditional GANs, namely fitting the generator’s architecture with residual blocks and experimenting with cycle consistency approaches. We studied the impact of these changes on visual verisimilarity and how an U-Net segmentation model is affected by the usage of synthetic data. We achieved sufficiently realistic-looking breast MRI images and maintained a stable segmentation score even when completely replacing the dataset with the synthetic set. Our results were promising, especially when concerning to Pix2PixHD and Residual CycleGAN architectures.

No abstract

No abstract

No abstract