PSENet: Psoriasis Severity Evaluation Network

Authors: Yi Li, Zhe Wu, Shuang Zhao, Xian Wu, Yehong Kuang, Yangtian Yan, Shen Ge, Kai Wang, Wei Fan, Xiang Chen, Yong Wang800-807

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To train and evaluate PSENet, we work with professional dermatologists from a top hospital and spend years in building a golden dataset. The experimental results show that PSENet can achieve the mean absolute error of 2.21 and the accuracy of 77.87% in pair comparison, outperforming baseline methods.
Researcher Affiliation Collaboration Yi Li,1 Zhe Wu,1,3 Shuang Zhao,2 Xian Wu,3 Yehong Kuang,2 Yangtian Yan,3 Shen Ge,3 Kai Wang,3 Wei Fan,3 Xiang Chen,2 Yong Wang1 1School of Automation, Central South University 2Department of Dermatology, Xiangya Hospital, Central South University 3Tencent Medical AI Lab {liyi1002, wuzhe950818, shuangxy, ywang}@csu.edu.cn, {yh 927, chenxiangck}@126.com {kevinxwu, yangtianyan, shenge, ironswang, davidwfan}@tencent.com
Pseudocode No The paper provides architectural diagrams and textual descriptions of the model, but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statement or link indicating that the source code for the described methodology is open-source or publicly available.
Open Datasets No In this study, we tracked the entire treatment processes of 1,787 Psoriasis patients and built a dataset consisting of 5,205 images where the longest recorded period is 15 months with 6 visits. The labels in this dataset include the severity scores which are annotated by 11 professional dermatologists (9 professors and 2 attending physicians), and the locations of skin lesions which are generated by our pretrained detection model. The paper does not provide concrete access information (link, DOI, or formal citation) for this dataset, implying it is not publicly available.
Dataset Splits Yes We split our data into 5 folds using individual patient as the smallest unit. The model is trained on 4 folds and evaluated on the held-out fold. All reported results are the average on 5 different validation folds.
Hardware Specification No The paper does not provide any specific details regarding the hardware (e.g., GPU/CPU models, memory, or processing units) used for running the experiments.
Software Dependencies No The paper mentions software components like ImageNet pre-trained weights, Adam optimizer, ResNet-50, GoogLeNet-v2, and Mask R-CNN, but it does not specify explicit version numbers for these software dependencies.
Experiment Setup Yes The model is initialized with weights pre-trained on Image Net (Deng et al. 2009), and Adam optimizer (Kingma and Ba 2014) has been used where the initial learning rate is 0.0001 and the weight decay is 0.0002. ... L = μ Lloc + ν Lsia + ξ Lreg where μ = 1.0, ν = 0.2, and ξ = 0.2 based on our experiments.