Learning to diagnose from scratch by exploiting dependencies among labels

By L. Yao, Eric Poblenz, Dmitry Dagunts, Ben Covington, Devon Bernard, Kevin Lyman. 2017

374 citations recorded by the source corpus.

The field of medical diagnostics contains a wealth of challenges which closely resemble classical machine learning problems; practical constraints, however, complicate the translation of these endpoints naively into classical architectures. Many tasks in radiology, for example, are largely problems of multi-label classification wherein medical images are interpreted to indicate multiple present or suspected pathologies. Clinical settings drive the necessity for high accuracy simultaneously across a multitude of pathological outcomes and greatly limit the utility of tools which consider only a subset. This issue is exacerbated by a general scarcity of training data and maximizes the need to extract clinically relevant features from available samples -- ideally without the use of pre-trained models which may carry forward undesirable biases from tangentially related tasks. We present and evaluate a partial solution to these constraints in using LSTMs to leverage interdependencies among target labels in predicting 14 pathologic patterns from chest x-rays and establish state of the art results on the largest publicly available chest x-ray dataset from the NIH without pre-training. Furthermore, we propose and discuss alternative evaluation metrics and their relevance in clinical practice.

Subfields: Artificial Intelligence, Radiology, Nuclear Medicine and Imaging

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 Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Deep learning predicts hip fracture using confounding patient and healthcare variables (2018, neutral): “st test test test test No. radiographs 17,587 5,970 416 405 416 411 No. patients 6,768 2,256 275 252 217 186 No. scanners 11 11 10 9 8 6 No. scanner manufacturers 4 4 4 4 4 4 Age, mean (SD), years 61 (22) 61 (22) 67 (24) 75 (20) 75 (21) 74 (19) Female frequency, No. (%) 11,647 (66) 3,873 (65) 260 (6”
CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning (2017, methodology): “%) sets, following previous work on ChestX-ray14 (Wang et al., 2017; Yao et al., 2017).”
CheXclusion: Fairness gaps in deep chest X-ray classifiers (2020, background): “Limiting a set of well-defined labels is hard due to the complex pattern between disease [48].”
Comparison of Deep Learning Approaches for Multi-Label Chest X-Ray Classification (2018, methodology): “In the work of Yao et al. 9 , they presented a combination of a CNN and a recurrent neural network to exploit label dependencies.”
The Effect of Image Resolution on Deep Learning in Radiography. (2020, methodology): “We pointedly moved away from some of the training paradigms of other studies (1,3,4) by training binary networks on dataset subsamples instead of multilabel models on the full dataset and, as a consequence, disinvite direct performance comparisons, because the validation set of images is not…”
Diagnose like a Radiologist: Attention Guided Convolutional Neural Network for Thorax Disease Classification (2018, methodology): “Yao et al. [19] and Kumar et al. [7] classify the chest X-ray images by investigating the potential dependencies among the labels from the aspect of multi-label problems.”
Deep-learning framework to detect lung abnormality - A study with chest X-Ray and lung CT scan images (2020, background): “X-Ray based assessment yields a very low accuracy due to the poor visibility of the abnormal section in radiographs compared to other modalities [19-22].”
Computer-aided detection in chest radiography based on artificial intelligence: a survey (2018, methodology): “LSTM + DENSENET [107] AUC Respectively, 77.”
An Efficient Deep Learning Approach to Pneumonia Classification in Healthcare (2019, methodology): “Statistical dependency between labels [19] was studied to arrive at more precise predictions, thereby outperforming other techniques on given 13 images selected from 14 classes [20].”