This repository contains code for the submission of NITS_AS21 team in the Diabetic Foot Ulcer Grand Challenge 2021 , which is organized as part of MICCAI 2021.
The DFUC21 challenge is accessible at https://dfu-2021.grand-challenge.org/
The MICCAI 2021 website can be accessed at https://www.miccai2021.org/en/
In this work, a Prediction Level Ensembling Approach is proposed which combines the predicitve capabilities of 2 Deep CNN models (Pretrained EfficientNetB0 and Pretrained DenseNet121).
The proposed approach has been evaluated on the held out validation set, the results for which are mentioned below:
"bothAUC": 0.9089912280701755,
"noneAUC": 0.8111724723874257,
"Accuracy": 0.594,
"macroAUC": 0.8439229430214277,
"microAUC": 0.8757613333333332,
"BothRecall": 0.3409090909090909,
"NoneRecall": 0.8084112149532711,
"MacroRecall": 0.5318968899751353,
"BothF1-Score": 0.38461538461538464,
"NoneF1-Score": 0.7004048582995951,
"infectionAUC": 0.7404431104871634,
"ischaemiaAUC": 0.9103780068728522,
"BothPrecision": 0.4411764705882353,
"MacroF1-Score": 0.48613309637572144,
"NonePrecision": 0.6178571428571429,
"MacroPrecision": 0.48970035382574084,
"InfectionRecall": 0.44493392070484583,
"IschaemiaRecall": 0.5333333333333333,
"InfectionF1-Score": 0.532981530343008,
"IschaemiaF1-Score": 0.326530612244898,
"InfectionPrecision": 0.6644736842105263,
"IschaemiaPrecision": 0.23529411764705882,
"WeightedAverageRecall": 0.594,
"WeightedAverageF1-Score": 0.5853889663414532,
"WeightedAveragePrecision": 0.6119962627156126