The seven key challenges for the future of computer-aided diagnosis in medicine

doi:10.1016/j.ijmedinf.2019.06.017

International Journal of Medical Informatics

Volume 129, September 2019, Pages 413-422

https://doi.org/10.1016/j.ijmedinf.2019.06.017 Get rights and content

Highlights

•
Proposed a new and more flexible approach (the Dynamic PRISMA Approach) for doing systematic literature reviews.
•
Grouped the challenges that CAD in medicine faces today into seven key categories and analyzed the relations among them.
•
The proposed grouping of the CAD challenges can help coordinate progress in CAD and elevate it to the next level.
•
Various data mining or machine learning techniques can greatly help dealing with most of these challenges.
•
A close collaboration between practitioners and researchers in medicine and computer science is crucial for meeting these challenges.

Abstract

Background

Computer-aided diagnosis (CAD) can assist physicians in effective and efficient diagnostic decision-making. CAD systems are currently essential tools in some areas of clinical practice. In addition, it is one of the established fields of study in the interface of medicine and computer science. There are, however, still some critical challenges that CAD systems face.

Methods

This paper first describes a new literature review protocol, the Dynamic PRISMA approach based on the well-known PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) approach. This new approach enhances the traditional approach by integrating a feedback mechanism module. As a result of the literature review, this paper identifies seven major challenges that occur today in CAD and inhibit the next major developments.

Results

The seven challenges described in this paper involve some technical weaknesses in the interface of medicine and computer science. These challenges are related to various algorithmic limitations, the difficulty of medical professionals to adopt new systems, problems when dealing with patient data, and the lack of guidelines and standardization regarding many aspects of CAD. This paper also describes some of the recent research developments towards these challenges.

Conclusion

If these seven key challenges are addressed properly, then the ways for dealing with them will become the R&D pillars needed to bring CAD to the next level. This would require additional well-coordinated collaboration between researchers and practitioners in the fields of medicine and computer science.

Introduction

The motivation for the present study became apparent after a recent systematic survey that the authors did on the subject of computer-aided diagnosis (CAD) in medicine regarding its past and current developments [1]. The first study indicated that there are some challenges still facing CAD development that seriously inhibit its progress. These challenges have to be understood well in order to bring CAD to the next level. By addressing these challenges, it is possible to eliminate existing barriers in CAD application and development.

Fig. 1 presents a brief history of the key advances in this area [2,57,[102], [103], [104], [105], [106], [107],109,110,112,113]. At the beginning of the early studies in CAD, researchers were hoping to develop entirely automatic computer-aided diagnostic systems. However, thanks to the better understanding reached in the early 1970s [2] of what computers can do or cannot do, this attitude has sifted. Currently, CAD is considered more as a diagnostic aid tool for physicians, as opposed to a completely automatic system for making medical diagnoses [3].

As CAD spans in the interface of medicine and computer science (see also Fig. 2), the challenges that CAD faces today are also related to these two constituent fields. Therefore, if progress in one of these fields occurs in relation to CAD, it is very likely that the other constituent field will benefit as well.

Section snippets

Methods

The initial literature search protocol followed was the well-accepted PRISMA approach [4]. The traditional PRISMA approach comes with some variants depending on the goal of the study (see also www.prisma-statement.org). However, existing PRISMA approaches are static, as they assume that the researchers already have a comprehensive understanding of all the issues involved (and hence the relevant keywords).

The database under consideration for the initial search was PubMed. The keywords / phrases,

Results and Discussion

The literature searches for CAD related publications have uncovered a number of challenges that modern CAD systems face. A typical CAD system, like most standard data mining systems, usually follows a four-step approach [6]. These steps are ordered as follows: data acquisition, data pre-processing, main processing, and visualization / presentation of the results. Moreover, the main processing step of CAD systems often involves three sub-steps. These sub-steps are as follows: data segmentation,

Conclusions

CAD systems play a significant role assisting physicians making objective and effective diagnostic decisions. This paper introduced a new literature review protocol, called the dynamic PRISMA approach. This approach is an enhancement of the traditional PRISMA approach used in such cases. Next it identified seven key challenges that CAD systems face today. This paper also described recent research efforts towards these seven challenges as well as it identified areas where future work has the

Contributors

The two authors conceived and designed the study. They drafted the manuscript. Both authors critically revised the work for important intellectual content, and approved the final version of the manuscript for publication.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

The originality of the submitted manuscript

This is an original paper and has not been submitted or presented anywhere else.

We also declare that there is no case of plagiarism in this manuscript.

Conflict of Interest Statement

There is no case of conflict of interest with any of the two co-authors of this manuscript.

Acknowledgments

The authors of this paper are very appreciative to the Editor-in-Chief Professor Heimar de Fátima Marin and the two anonymous reviewers the comments of which were valuable in clarifying and enhancing the original version of this paper.

References (114)

K. Doi
Computer-aided diagnosis in medical imaging: historical review, current status and future potential
Comput. Med. Imaging Graph.
(2007)
K. Kamnitsas et al.
Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation
Med. Image Anal.
(2017)
R. Togo et al.
Cardiac sarcoidosis classification with deep convolutional neural network-based features using polar maps
Comput. Biol. Med.
(2019)
Z. Zhu et al.
Three-dimensional Gabor feature extraction for hyperspectral imagery classification using a memetic framework
Inf. Sci. (Ny)
(2015)
S. Pereira et al.
Automatic brain tissue segmentation in MR images using random forests and conditional random fields
J. Neurosci. Methods
(2016)
B.D. Gallas et al.
Evaluating imaging and computer-aided detection and diagnosis devices at the FDA
Acad. Radiol.
(2012)
E.E. Bron et al.
Standardized evaluation of algorithms for computer-aided diagnosis of dementia based on structural MRI: the CADDementia challenge
NeuroImage
(2015)
D. Regge et al.
CAD: how it works, how to use it, performance
Eur. J. Radiol.
(2013)
L.A. Zadeh
Fuzzy sets
Inf. Control.
(1965)
K. Atanassov
Intuitionistic fuzzy sets
Fuzzy Sets Syst.
(1986)

E. Triantaphyllou et al.

Development and evaluation of five fuzzy multiattribute decision-making methods

Int. J. Approx. Reason.

(1996)

B. Kovalerchuk et al.

Fuzzy logic in computer-aided breast cancer diagnosis: analysis of lobulation

Artif. Intell. Med.

(1997)

K.P. Lin et al.

Intuitionistic fuzzy C-regression by using least squares support vector regression

Expert Syst. Appl.

(2016)

R.J. Kuo et al.

A hybrid metaheuristic and kernel intuitionistic fuzzy c-means algorithm for cluster analysis

Appl. Soft Comput.

(2018)

I.K. Vlachos et al.

Intuitionistic fuzzy information–applications to pattern recognition

Pattern Recognit. Lett.

(2007)

Y. Chen et al.

IBM Watson: how cognitive computing can be applied to big data challenges in life sciences research

Clin. Ther.

(2016)

R.M. Summers

Evaluation of computer-aided detection devices

Acad. Radiol.

(2012)

J. Yanase et al.

A systematic survey of computer-aided diagnosis in medicine: past and present developments

Expert Syst. Appl.

(2019)

R.M. Karp

Reducibility among combinatorial problems

Complexity of Computer Computations

(1972)

D. Moher et al.

Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement

Ann. Intern. Med.

(2009)

J. Beall

What I learned from predatory publishers

Biochem. Med. (Zagreb)

(2017)

E. Triantaphyllou

(2010)

M.G. Weiner et al.

Toward reuse of clinical data for research and quality improvement: the end of the beginning?

Ann. Intern. Med.

(2009)

J.S. Adelman et al.

A national survey assessing the number of records allowed open in electronic health records at hospitals and ambulatory sites

J. Am. Med. Inform. Assoc.

(2017)

J. Pagliery

Hospital network hacked

4.5 Million Records Stolen

(2014)

G. Hripcsak et al.

Next-generation phenotyping of electronic health records

J. Am. Med. Inform. Assoc.

(2013)

S. Bowman

Impact of electronic health record systems on information integrity: quality and safety implications

Perspect. Health Inf. Manag.

(2013)

N.G. Weiskopf et al.

Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research

J. Am. Med. Inform. Assoc.

(2013)

M.J. Santana et al.

Training clinicians in how to use patient-reported outcome measures in routine clinical practice

Qual. Life Res.

(2015)

K. Stroup et al.

A new EHR training curriculum and assessment for pediatric residents

Appl. Clin. Inform.

(2017)

A. Hardon et al.

How to Investigate the Use of Medicines by Consumers

(2004)

L.G. Valiant

A theory of the learnable

Commun. ACM

(1984)

C.J. Pannucci et al.

Identifying and avoiding bias in research

Plast. Reconstr. Surg.

(2010)

M. Sonka et al.

Image Processing, Analysis, and Machine Vision

(2014)

B. van Ginneken et al.

Computer-aided diagnosis: how to move from the laboratory to the clinic

Radiology

(2011)

D. Yang et al.

Automatic liver segmentation using an adversarial image-to-image network

International Conference on Medical Image Computing and Computer-Assisted Intervention

(2017)

A. Kumar et al.

Content-based medical image retrieval: a survey of applications to multidimensional and multimodality data

J. Digit. Imaging

(2013)

A. Kumar et al.

An ensemble of fine-tuned convolutional neural networks for medical image classification

IEEE J. Biomed. Health Inform.

(2016)

R. Girshick et al.

Rich feature hierarchies for accurate object detection and semantic segmentation

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

(2014)

J. Long et al.

Fully convolutional networks for semantic segmentation

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

(2015)

B. Romera-Paredes et al.

Recurrent instance segmentation

European Conference on Computer Vision, October Springer, Cham

(2016)

I. Goodfellow et al.

(2016)

Y. LeCun et al.

Deep learning

Nature

(2015)

W. Wu et al.

Brain tumor detection and segmentation in a CRF (conditional random fields) framework with pixel-pairwise affinity and superpixel-level features

Int. J. Comput. Assist. Radiol. Surg.

(2014)

P.F. Christ et al.

October. Automatic liver and lesion segmentation in CT using cascaded fully convolutional neural networks and 3D conditional random fields

International Conference on Medical Image Computing and Computer-Assisted Intervention

(2016)

H.C. Shin et al.

Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning

IEEE Trans. Med. Imaging

(2016)

K. He et al.

Mask r-cnn

Computer Vision (ICCV), 2017 IEEE International Conference, October, IEEE

(2017)

Y. Bar et al.

Chest pathology identification using deep feature selection with non-medical training

Comput. Methods Biomech. Biomed. Eng. Imaging Vis.

(2018)

R.L. Thomaz et al.

Feature extraction using convolutional neural network for classifying breast density in mammographic images

Medical Imaging 2017: Computer-Aided Diagnosis (Vol. 10134, p. 101342M)

(2017)

P. Moeskops et al.

Automatic segmentation of MR brain images with a convolutional neural network

IEEE Trans. Med. Imaging

(2016)

Cited by (46)

Automated image label extraction from radiology reports — A review
2024, Artificial Intelligence in Medicine
Machine Learning models need large amounts of annotated data for training. In the field of medical imaging, labeled data is especially difficult to obtain because the annotations have to be performed by qualified physicians. Natural Language Processing (NLP) tools can be applied to radiology reports to extract labels for medical images automatically. Compared to manual labeling, this approach requires smaller annotation efforts and can therefore facilitate the creation of labeled medical image data sets. In this article, we summarize the literature on this topic spanning from 2013 to 2023, starting with a meta-analysis of the included articles, followed by a qualitative and quantitative systematization of the results. Overall, we found four types of studies on the extraction of labels from radiology reports: those describing systems based on symbolic NLP, statistical NLP, neural NLP, and those describing systems combining or comparing two or more of the latter. Despite the large variety of existing approaches, there is still room for further improvement. This work can contribute to the development of new techniques or the improvement of existing ones.
Detection of dental restorations using no-code artificial intelligence
2023, Journal of Dentistry
The purpose of this study was to utilize a no-code computer vision platform to develop, train, and evaluate a model specifically designed for segmenting dental restorations on panoramic radiographs.
One hundred anonymized panoramic radiographs were selected for this study. Accurate labeling of dental restorations was performed by calibrated dental faculty and students, with subsequent final review by an oral radiologist. The radiographs were automatically split within the platform into training (70 %), development (20 %), and testing (10 %) subgroups. The model was trained for 40 epochs using a medium model size. Data augmentation techniques available within the platform, namely horizontal and vertical flip, were utilized on the training set to improve the model's predictions. Post-training, the model was tested for independent predictions. The model's diagnostic validity was assessed through the calculation of sensitivity, specificity, accuracy, precision, F1-score by pixel and by tooth, and by ROC-AUC.
A total of 1,108 restorations were labeled on 960 teeth. At a confidence threshold of 0.95, the model achieved 86.64 % sensitivity, 99.78 % specificity, 99.63 % accuracy, 82.4 % precision and an F1-score of 0.844 by pixel. The model achieved 98.34 % sensitivity, 98.13 % specificity, 98.21 % accuracy, 98.85 % precision and an F1-score of 0.98 by tooth. ROC curve showed high performance with an AUC of 0.978.
The no-code computer vision platform used in this study accurately detected dental restorations on panoramic radiographs. However, further research and validation are required to evaluate the performance of no-code platforms on larger and more diverse datasets, as well as for other detection and segmentation tasks.
The advent of no-code computer vision holds significant promise in dentistry and dental research by eliminating the requirement for coding skills, democratizing access to artificial intelligence tools, and potentially revolutionizing dental diagnostics.
Big data in healthcare: Conceptual network structure, key challenges and opportunities
2023, Digital Communications and Networks
Big data is a concept that deals with large or complex data sets by using data analysis tools (e.g., data mining, machine learning) to analyze information extracted from several sources systematically. Big data has attracted wide attention from academia, for example, in supporting patients and health professionals by improving the accuracy of decision-making, diagnosis and disease prediction. This research aimed to perform a Bibliometric Performance and Network Analysis (BPNA) supported by a Scoping Review (SR) to depict the strategic themes, thematic evolution structure, main challenges and opportunities related to the concept of big data applied in the healthcare sector. With this goal in mind, 4857 documents from the Web of Science covering the period between 2009 to June 2020 were analyzed with the support of SciMAT software. The bibliometric performance showed the number of publications and citations over time, scientific productivity and the geographic distribution of publications and research fields. The strategic diagram yielded 20 clusters and their relative importance in terms of centrality and density. The thematic evolution structure presented the most important themes and how it changes over time. Lastly, we presented the main challenges and future opportunities of big data in healthcare.
Using feature maps to unpack the CNN ‘Black box’ theory with two medical datasets of different modality
2023, Intelligent Systems with Applications
Convolutional neural networks (CNNs) have been established for a comprehensive range of computer vision problems across several benchmarks. Visualization and analysis of feature maps generated by convolutional layers can be an effective approach to explore the hidden and complex characteristic of a CNN model. Convolutional layers provide diverse feature maps however, the extent of this diversity needs to be explored. This research attempts to provide five insights of the ‘Black box’ mechanism of CNNs, using skin cancer dermoscopy and lung scan computed tomography (CT) Scan datasets by statistically analyzing layer by layer (three convolutional layers) feature maps using 17 geometrical and 6 intensity-based features to determine the characteristics and level of diversity. Significance and difference of the feature maps layer by layer, black feature maps analysis, difference of the feature maps to each other and to the original image, variations among the feature maps when running the model multiple times and inter-class variation among the feature maps for different iteration are explored. Various statistical methods including T-test, analysis of variance (ANOVA), mean, median, mean squared error (MSE), peak signal to noise ratio (PSNR), structural similarity index (SSIM), root mean squared error (RMSE), dice similarity score (DSC), universal image quality index (UQI) and Spectral angle mapper (SAM) are employed. Experimental results show that for the skin cancer dermoscopy dataset, a large number of black feature maps are produced (20–60%) while the proportion of black feature maps for the CT Scan dataset is comparatively low (2–20%). This demonstrates that for different datasets, feature maps with diverse characteristics can be produced. The layer by layer differences between the feature maps is evaluated using T-tests and ANOVA for seventeen geometrical features and six intensity-based features. For both datasets across most of the geometrical features and across most of the intensity-based features a significant diversity can be observed. The difference of the feature maps to each other and to the original image is quite high, with MSE values for the dermoscopy and CT Scan datasets in the range of 1860–31,399 and 171–6089, respectively, PSNR 3–15 and 10–25, SSIM values of 0.01–0.84 and 0.3–0.81, RMSE values of 0.81–1 and 0.21–1, DSC values of 0.37–0.53 and 0.47–0.75, UQI values of 0.02–0.86 and 0.01–0.88 and SAM values of 0.12–1.53 and 0.19–1.55 for the dermoscopy and CT Scan datasets respectively. When running the model multiple times (three iterations), a notable iteration by iteration diversity is found in terms of mean, median, maximum and minimum values for most of the geometrical features. The inter-class variation among the feature maps for different iterations and layers are evaluated based on the F-value of the ANOVA test. For the dermoscopy dataset, the highest mean F-value is found for layer 1 and iteration 3 while for the CT scan dataset the highest mean F-value is found for layer 3 and iteration 3 indicating that for these feature maps the highest inter-class dissimilarity is generated. The findings of this study may aid in exploring the complex mechanism of convolutional layers, kernels and feature maps.
Disease Localization and Severity Assessment in Chest X-Ray Images using Multi-Stage Superpixels Classification
2022, Computer Methods and Programs in Biomedicine
Citation Excerpt :
While subdivided the lung fields, the number of non-infected superpixels is significantly larger than that of infected superpixels, which substantially affects the performance. Moreover, to train a generalized model, a balanced dataset is required [65]. The results of this study demonstrate that infected region in CXR images can be efficiently identified using multistage superpixel classification approach.
Chest X-ray (CXR) is a non-invasive imaging modality used in the prognosis and management of chronic lung disorders like tuberculosis (TB), pneumonia, coronavirus disease (COVID-19), etc. The radiomic features associated with different disease manifestations assist in detection, localization, and grading the severity of infected lung regions. The majority of the existing computer-aided diagnosis (CAD) system used these features for the classification task, and only a few works have been dedicated to disease-localization and severity scoring. Moreover, the existing deep learning approaches use class activation map and Saliency map, which generate a rough localization. This study aims to generate a compact disease boundary, infection map, and grade the infection severity using proposed multistage superpixel classification-based disease localization and severity assessment framework.
: The proposed method uses a simple linear iterative clustering (SLIC) technique to subdivide the lung field into small superpixels. Initially, the different radiomic texture and proposed shape features are extracted and combined to train different benchmark classifiers in a multistage framework. Subsequently, the predicted class labels are used to generate an infection map, mark disease boundary, and grade the infection severity. The performance is evaluated using a publicly available Montgomery dataset and validated using Friedman average ranking and Holm and Nemenyi post-hoc procedures.
: The proposed multistage classification approach achieved accuracy (ACC)= 95.52%, F-Measure (FM)= 95.48%, area under the curve (AUC)= 0.955 for Stage-I and ACC=85.35%, FM=85.20%, AUC=0.853 for Stage-II using calibration dataset and ACC = 93.41%, FM = 95.32%, AUC = 0.936 for Stage-I and ACC = 84.02%, FM = 71.01%, AUC = 0.795 for Stage-II using validation dataset. Also, the model has demonstrated the average Jaccard Index (JI) of 0.82 and Pearson's correlation coefficient (r) of 0.9589.
: The obtained classification results using calibration and validation dataset confirms the promising performance of the proposed framework. Also, the average JI shows promising potential to localize the disease, and better agreement between radiologist score and predicted severity score (r) confirms the robustness of the method. Finally, the statistical test justified the significance of the obtained results.
Artificial neural network-based approaches for computer-aided disease diagnosis and treatment
2022, Cognitive and Soft Computing Techniques for the Analysis of Healthcare Data
The adoption of computer-aided diagnosis and treatment systems based on different types of artificial neural networks (ANNs) is already a reality in several hospital and ambulatory premises. This chapter aims to present a discussion focused on the challenges and trends of adopting these computerized systems, highlighting solutions based on different types and approaches of ANN, more specifically, feed-forward, recurrent, and deep convolutional architectures. One section is focused on the application of AI/ANN solutions to support cardiology in different applications, such as the classification of the heart structure and functional behavior based on echocardiography images; the automatic analysis of the heart electric activity based on ECG signals; and the diagnosis support of angiogram images during surgical interventions. Finally, a case study is presented based on the application of a deep learning convolutional network together with a recent technique called transfer learning to detect brain tumors using an MRI images data set. According to the findings, the model has a high degree of specificity (precision of 0.93 and recall of 0.94 for images with no brain tumor) and can be used as a screening tool for images that do not contain a brain tumor. The f1-score for images with brain tumor was 0.93. The results achieved are very promising and the proposed solution may be considered to be used as a computer-aided diagnosis tool based on deep learning convolutional neural networks. Future works will consider other techniques and compare them with the one presented here. With the comprehensive approach and overview of multiple applications, it is valid to conclude that computer-aided diagnosis and treatment systems are important tools to be considered today and will be an essential part of the trend of personalized medicine over the coming years.

View all citing articles on Scopus

View full text

The seven key challenges for the future of computer-aided diagnosis in medicine

Highlights

Abstract

Background

Methods

Results

Conclusion

Introduction

Section snippets

Methods

Results and Discussion

Conclusions

Contributors

Funding

The originality of the submitted manuscript

Conflict of Interest Statement

Acknowledgments

Comput. Med. Imaging Graph.

Med. Image Anal.

Comput. Biol. Med.

Inf. Sci. (Ny)

J. Neurosci. Methods

Acad. Radiol.

NeuroImage

Eur. J. Radiol.

Inf. Control.

Fuzzy Sets Syst.

Int. J. Approx. Reason.

Artif. Intell. Med.

Expert Syst. Appl.

Appl. Soft Comput.

Pattern Recognit. Lett.

Clin. Ther.

Acad. Radiol.

A systematic survey of computer-aided diagnosis in medicine: past and present developments

Expert Syst. Appl.

Reducibility among combinatorial problems

Complexity of Computer Computations

Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement

Ann. Intern. Med.

What I learned from predatory publishers

Biochem. Med. (Zagreb)

Toward reuse of clinical data for research and quality improvement: the end of the beginning?

Ann. Intern. Med.

A national survey assessing the number of records allowed open in electronic health records at hospitals and ambulatory sites

J. Am. Med. Inform. Assoc.

Hospital network hacked

4.5 Million Records Stolen

Next-generation phenotyping of electronic health records

J. Am. Med. Inform. Assoc.

Impact of electronic health record systems on information integrity: quality and safety implications

Perspect. Health Inf. Manag.

Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research

J. Am. Med. Inform. Assoc.

Training clinicians in how to use patient-reported outcome measures in routine clinical practice

Qual. Life Res.

A new EHR training curriculum and assessment for pediatric residents

Appl. Clin. Inform.

How to Investigate the Use of Medicines by Consumers

A theory of the learnable

Commun. ACM

Identifying and avoiding bias in research

Plast. Reconstr. Surg.

Image Processing, Analysis, and Machine Vision

Computer-aided diagnosis: how to move from the laboratory to the clinic

Radiology

Automatic liver segmentation using an adversarial image-to-image network

International Conference on Medical Image Computing and Computer-Assisted Intervention

Content-based medical image retrieval: a survey of applications to multidimensional and multimodality data

J. Digit. Imaging

An ensemble of fine-tuned convolutional neural networks for medical image classification

IEEE J. Biomed. Health Inform.

Rich feature hierarchies for accurate object detection and semantic segmentation

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Fully convolutional networks for semantic segmentation

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Recurrent instance segmentation

European Conference on Computer Vision, October Springer, Cham

Deep learning

Nature