search

cwmok / DIRAC

This is the official Pytorch implementation of "Unsupervised Deformable Image Registration with Absent Correspondences in Pre-operative and Post-Recurrence Brain Tumor MRI Scans" (MICCAI 2022), written by Tony C. W. Mok and Albert C. S. Chung.
MIT License
37 stars 1 forks source link

issue about the calculation of metric #15

Closed 18370029656 closed 9 months ago

18370029656 commented 10 months ago

Hello, I'm sorry to bother you. I have some questions about the calculation of the evaluation metrics mentioned in your paper, and I hope to receive your answer. When calculating the metrics, you divided them into two sets: near the tumor area and far from the tumor area. Does the near tumor area include the tumor area? If the tumor area is not included, why are the metrics values near the tumor area generally much lower than those far away from the tumor area when you have already ruled out the influence of the tumor area

cwmok commented 10 months ago

Hi @18370029656, the whole tumor is included in the near tumor area as defined in the paper.

18370029656 commented 10 months ago

ok,thanks for your quick reply!

18370029656 commented 9 months ago

Hello, may I ask if the near tumor region within 30mm from the tumor region is calculated from the tumor center or from the tumor boundary when dividing the area?

------------------ 原始邮件 ------------------ 发件人: "cwmok/DIRAC" @.>; 发送时间: 2024年1月8日(星期一) 晚上10:20 @.>; @.**@.>; 主题: Re: [cwmok/DIRAC] issue about the calculation of metric (Issue #15)

Hi @18370029656, the whole tumor is included in the near tumor area as defined in the paper.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

cwmok commented 9 months ago

@18370029656 It is computed from tumor boundary.

18370029656 commented 9 months ago

Okay, thanks for your quick reply!

------------------ 原始邮件 ------------------ 发件人: "cwmok/DIRAC" @.>; 发送时间: 2024年1月29日(星期一) 下午3:48 @.>; @.**@.>; 主题: Re: [cwmok/DIRAC] issue about the calculation of metric (Issue #15)

@18370029656 It is computed from tumor boundary.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

18370029656 commented 8 months ago

Hello, I have a little question for you about the calculation of evaluation indicators. When comparing VM with VM-CM, I found that in the group of near-tumour regions, the values of the metrics (dice, etc.) are higher in VM than in VM-CM, is this not reasonable? My understanding is that the VM method forces the alignment of the tumour images in the near tumour region because it does not consider the tumour region separately, whereas the VM-CM excludes the effect of the tumour region and does not take into account the similarity of the tumour region, so the metrics values will be lower. May I ask if my understanding is accurate? Looking forward to your reply!

------------------ 原始邮件 ------------------ 发件人: "123" @.>; 发送时间: 2024年1月29日(星期一) 下午3:56 @.>;

主题: 回复： [cwmok/DIRAC] issue about the calculation of metric (Issue #15)

Okay, thanks for your quick reply!

------------------ 原始邮件 ------------------ 发件人: "cwmok/DIRAC" @.>; 发送时间: 2024年1月29日(星期一) 下午3:48 @.>; @.**@.>; 主题: Re: [cwmok/DIRAC] issue about the calculation of metric (Issue #15)

@18370029656 It is computed from tumor boundary.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>