-
Teaching AI the Anatomy Behind the Scan: Addressing Anatomical Flaws in Medical Image Segmentation with Learnable Prior
Authors:
Young Seok Jeon,
Hongfei Yang,
Huazhu Fu,
Mengling Feng
Abstract:
Imposing key anatomical features, such as the number of organs, their shapes and relative positions, is crucial for building a robust multi-organ segmentation model. Current attempts to incorporate anatomical features include broadening the effective receptive field (ERF) size with data-intensive modules, or introducing anatomical constraints that scales poorly to multi-organ segmentation. We intr…
▽ More
Imposing key anatomical features, such as the number of organs, their shapes and relative positions, is crucial for building a robust multi-organ segmentation model. Current attempts to incorporate anatomical features include broadening the effective receptive field (ERF) size with data-intensive modules, or introducing anatomical constraints that scales poorly to multi-organ segmentation. We introduce a novel architecture called the Anatomy-Informed Cascaded Segmentation Network (AIC-Net). AIC-Net incorporates a learnable input termed "Anatomical Prior", which can be adapted to patient-specific anatomy using a differentiable spatial deformation. The deformed prior later guides decoder layers towards more anatomy-informed predictions. We repeat this process at a local patch level to enhance the representation of intricate objects, resulting in a cascaded network structure. AIC-Net is a general method that enhances any existing segmentation models to be more anatomy-aware. We have validated the performance of AIC-Net, with various backbones, on two multi-organ segmentation tasks: abdominal organs and vertebrae. For each respective task, our benchmarks demonstrate improved dice score and Hausdorff distance.
△ Less
Submitted 26 August, 2024; v1 submitted 27 March, 2024;
originally announced March 2024.
-
Exploring the magnetic properties of individual barcode nanowires using wide-field diamond microscopy
Authors:
Jungbae Yoon,
Jun Hwan Moon,
Jugyeong Jeong,
Yu Jin Kim,
Kihwan Kim,
Hee Seong Kang,
Yoo Sang Jeon,
Eunsoo Oh,
Sun Hwa Lee,
Kihoon Han,
Dongmin Lee,
Chul-Ho Lee,
Young Keun Kim,
Donghun Lee
Abstract:
Barcode magnetic nanowires typically comprise a multilayer magnetic structure in a single body with more than one segment type. Interestingly, owing to selective functionalization and novel interactions between the layers, barcode magnetic nanowires have attracted significant attention, particularly in the field of bioengineering. However, an analysis of their magnetic properties at the individual…
▽ More
Barcode magnetic nanowires typically comprise a multilayer magnetic structure in a single body with more than one segment type. Interestingly, owing to selective functionalization and novel interactions between the layers, barcode magnetic nanowires have attracted significant attention, particularly in the field of bioengineering. However, an analysis of their magnetic properties at the individual nanowire level remains challenging. With this background, herein, we investigated the characterization of magnetic nanowires at room temperature under ambient conditions based on magnetic images obtained via wide-field quantum microscopy with nitrogen-vacancy centers in diamond. Consequently, we could extract critical magnetic properties, such as the saturation magnetization and coercivity, of single nanowires by comparing the experimental results with those of micromagnetic simulations. This study opens up the possibility for a versatile characterization method suited to individual magnetic nanowires.
△ Less
Submitted 21 February, 2023;
originally announced February 2023.
-
Interwire and Intrawire Magnetostatic Interactions in Fe-Au Barcode Nanowires with Alternating Ferromagnetically Strong and Weak Segments
Authors:
Aleksei Yu. Samardak,
Yoo Sang Jeon,
Vadim Yu. Samardak,
Alexey G. Kozlov,
Kirill A. Rogachev,
Alexey V. Ognev,
Eunjin Jeong,
Gyu Won Kim,
Min Jun Ko,
Alexander S. Samardak,
Young Keun Kim
Abstract:
Metallic barcode nanowires (BNWs) composed of repeating heterogeneous segments fabricated by template-assisted electrodeposition can offer extended functionality in magnetic, electrical, mechanical, and biomedical applications. We can consider such nanostructures as a three-dimensional system of magnetically interacting elements with magnetic behavior strongly affected by complex magnetostatic int…
▽ More
Metallic barcode nanowires (BNWs) composed of repeating heterogeneous segments fabricated by template-assisted electrodeposition can offer extended functionality in magnetic, electrical, mechanical, and biomedical applications. We can consider such nanostructures as a three-dimensional system of magnetically interacting elements with magnetic behavior strongly affected by complex magnetostatic interactions. This study discusses the influence of geometrical parameters of segments on the character of their interactions and the overall magnetic behavior of the array of BNWs having alternating magnetization. By controlling the applied current densities and the elapsed time in the electrodeposition, we regulate the dimension of the Fe-Au BNWs. We show that the Fe and Au segments are made of Fe-Au alloys with high and low magnetization. With the help of micromagnetic simulations, we discover and analyze the three types of magnetostatic interactions in the BNW arrays. As a result, we demonstrate that the dominating type of interaction depends on the geometric parameters of the Fe and Au segments and the interwire and intrawire distances.
△ Less
Submitted 13 September, 2022;
originally announced September 2022.
-
FCSN: Global Context Aware Segmentation by Learning the Fourier Coefficients of Objects in Medical Images
Authors:
Young Seok Jeon,
Hongfei Yang,
Mengling Feng
Abstract:
The encoder-decoder model is a commonly used Deep Neural Network (DNN) model for medical image segmentation. Conventional encoder-decoder models make pixel-wise predictions focusing heavily on local patterns around the pixel. This makes it challenging to give segmentation that preserves the object's shape and topology, which often requires an understanding of the global context of the object. In t…
▽ More
The encoder-decoder model is a commonly used Deep Neural Network (DNN) model for medical image segmentation. Conventional encoder-decoder models make pixel-wise predictions focusing heavily on local patterns around the pixel. This makes it challenging to give segmentation that preserves the object's shape and topology, which often requires an understanding of the global context of the object. In this work, we propose a Fourier Coefficient Segmentation Network~(FCSN) -- a novel DNN-based model that segments an object by learning the complex Fourier coefficients of the object's masks. The Fourier coefficients are calculated by integrating over the whole contour. Therefore, for our model to make a precise estimation of the coefficients, the model is motivated to incorporate the global context of the object, leading to a more accurate segmentation of the object's shape. This global context awareness also makes our model robust to unseen local perturbations during inference, such as additive noise or motion blur that are prevalent in medical images. When FCSN is compared with other state-of-the-art models (UNet+, DeepLabV3+, UNETR) on 3 medical image segmentation tasks (ISIC\_2018, RIM\_CUP, RIM\_DISC), FCSN attains significantly lower Hausdorff scores of 19.14 (6\%), 17.42 (6\%), and 9.16 (14\%) on the 3 tasks, respectively. Moreover, FCSN is lightweight by discarding the decoder module, which incurs significant computational overhead. FCSN only requires 22.2M parameters, 82M and 10M fewer parameters than UNETR and DeepLabV3+. FCSN attains inference and training speeds of 1.6ms/img and 6.3ms/img, that is 8$\times$ and 3$\times$ faster than UNet and UNETR.
△ Less
Submitted 29 July, 2022;
originally announced July 2022.
