This is the official project page for our work on IVOCT lesion segmentation and detection.

Our lab page is available at lab page.

Our research has been accepted for publication in the IEEE Transactions on Medical Imaging (TMI) 2024 and presented at MICCAI 2023.

We collect a large-scale IVOCT dataset, including segmentation annotations and object detection annotations.

We plan to share the dataset in the future. If you need access to the dataset, please contact us via email.

PolarFormer: A Transformer-based Method for Multi-lesion Segmentation in Intravascular OCT

Fig 1: The architecture of our proposed model PolarFormer. Top: Our model consists of an encoder and a decoder. The encoder includes a three-layer downsampling CNN and our new Polar-Attention module. The decoder features a symmetrical three-layer upsampling structure. The encoder and decoder are connected by skip connections. Bottom: Details of our proposed Polar-Attention.

The paper on IVOCT lesion segmentation is available here: coming soon

The code for IVOCT lesion segmentation can be found here: coming soon

Vision Transformer Based Multi-class Lesion Detection in IVOCT

Fig 2: We proposed a Vision Transformer-based model, called G-Swin Transformer, which uses adjacent frames as input and leverages the temporary dimensional information inherent in IVOCT data.

The paper on IVOCT lesion detection is available here: paper

The code for IVOCT lesion detection can be found here: code