Fine-tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally

By Zongwei Zhou, Jae Y. Shin, Lei Zhang, Suryakanth R. Gurudu, Michael B. Gotway, Jianming Liang. Computer Vision and Pattern Recognition, 2017

409 citations recorded by the source corpus.

Intense interest in applying convolutional neural networks (CNNs) in biomedical image analysis is wide spread, but its success is impeded by the lack of large annotated datasets in biomedical imaging. Annotating biomedical images is not only tedious and time consuming, but also demanding of costly, specialty - oriented knowledge and skills, which are not easily accessible. To dramatically reduce annotation cost, this paper presents a novel method called AIFT (active, incremental fine-tuning) to naturally integrate active learning and transfer learning into a single framework. AIFT starts directly with a pre-trained CNN to seek worthy samples from the unannotated for annotation, and the (fine-tuned) CNN is further fine-tuned continuously by incorporating newly annotated samples in each iteration to enhance the CNNs performance incrementally. We have evaluated our method in three different biomedical imaging applications, demonstrating that the cost of annotation can be cut by at least half. This performance is attributed to the several advantages derived from the advanced active and incremental capability of our AIFT method.

Subfields: Artificial Intelligence, Biophysics, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/71cdc2345df62d60a1ed88333a6101ce669663b0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Biophysics

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

A Survey on Active Learning and Human-in-the-Loop Deep Learning for Medical Image Analysis (2019, methodology): “Their previous work Zhou et al. (2017) applies a similar but less featureful approach to several medical image analysis tasks to demonstrate equivalent performance can be reached with a heavily reduced training data-set.”
Class Boundary Exemplar Selection Based Incremental Learning for Automatic Target Recognition (2020, background): “Whether the sample is representative or not depends on some output indexes of the classiﬁcation model, such as classiﬁcation conﬁdence [33], probability output [21], entropy [20], predicting error [34], and interface-distance-based sampling [22].”
Deep learning in medical imaging and radiation therapy. (2018, methodology): “Zhou et al. 342 measured the uncertainty and diversity of candidate image samples using the CNN classification prediction values computed for all the image patches extracted from the candidate image.”
Polyp-PVT: Polyp Segmentation with Pyramid Vision Transformers (2021, background): “CL = CVC-CLINIC, EL = ETIS-LARIB, C6 = CVC-612, AM = ASU-MAYO [41], [42], ES = ENDOSCENE, DB = COLONDB, CV = CVC-VIDEOCLINICDB, C = COLON, ED = ENDOTECT 2020, KS = KVASIR-SEG, KCS = KVASIR CAPSULE-SEG, PRANET = SAME TO DATASETS USED IN PRANET [5], IS = IMAGE SEGMENTATION, VS = VIDEO SEGMENTATION, CF”
Learning Loss for Active Learning (2019, background): “Thus, a variety of hybrid strategies have been proposed [29, 59, 41, 56] for their specific tasks.”
FITAnnotator: A Flexible and Intelligent Text Annotation System (2021, methodology): “In particular, we employ a fused active learning method as a default strategy for evaluating, re-ranking and re-sampling data, which considers uncertainty and diversity at the same time (Zhou et al., 2017; Lutnick et al., 2019).”
Continuous Adaptation for Interactive Object Segmentation by Learning from Corrections (2019, background): “For example, when supervision for the first frame of a test video is available [43,55,13], or after annotating a subset of an image dataset [42,61].”
UNet++: A Nested U-Net Architecture for Medical Image Segmentation (2018, background): “colon polyp 7,379 224×224 RGB video ASU-Mayo [10,11]”
Attention Residual Learning for Skin Lesion Classification (2019, positive): “To apply deep learning techniques directly to skin lesion classiﬁcation, a straightforward solution is to collect more training data and ﬁne-tune a powerful pre-trained DCNN [9], [10], [11], [12].”