AUTOMATED LUNG CANCER DIAGNOSIS USING THREE-DIMENSIONAL CONVOLUTIONAL NEURAL NETWORKS

G.PÉREZ AND P. ARBELÁEZ

MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING

Abstract

Lung cancer is the deadliest cancer worldwide. It has been shown that early detection using low-dose computer tomography (LDCT) scans can reduce deaths caused by this disease. We present a general framework for the detection of lung cancer in chest LDCT images. Our method consists of a nodule detector trained on the LIDC-IDRI dataset followed by a cancer predictor trained on the Kaggle DSB 2017 dataset and evaluated on the IEEE International Symposium on Biomedical Imaging (ISBI) 2018 Lung Nodule Malignancy Prediction test set. Our candidate extraction approach is effective to produce accurate candidates with a recall of 99.6%. In addition, our false positive reduction stage classifies successfully the candidates and increases precision by a factor of 2000. Our cancer predictor obtained a ROC AUC of 0.913 and was ranked 1st place at the ISBI 2018 Lung Nodule Malignancy Prediction challenge.

Method

Results

METHODAP (%)
No FP reduction method
HOG + SVM
CNN 2D
CNN 2.5D
CNN 3D		0.0062
0.021
8.9
36.7
44.8
Table 1. Average precision obtained with different false positive reduction methods of the nodule detector.
METHODAUC ROC (%)
Validation set
5-way Siamese cancer predictor
+ 3D mask subtraction (X = 0.82)
Test set
5-way Siamese cancer predictor
+ 3D mask subtraction (X = 0.82)		91.5


91.3*
Table 2. AUC of the ROC curve obtained on the validation and test set with and without post-processing.
* ISBI 2018 lung cancer challenge results can be found here.

Figure 2. Comparison between false positive reduction methods. HOG+SVM and best networks using convolutions in 2D, 2.5D (2D convolutions over 3D candidates) and 3D.

Universidad de los Andes | Monitored by Mineducación
Recognition as University: Decree 1297 of May 30th, 1964.
Recognition as legal entity: Resolution 28 of February 23, 1949 Minjusticia.

© Universidad de los Andes. All rights reserved.