Is there a role for image science in the brave new world of artificial intelligence?

Barrett, Harrison H.

doi:10.1117/1.jmi.7.1.012702

Cited by 5 publications

(3 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides understanding perception in medical images, our results could benefit multiple areas such as artificial intelligence, machine learning and computer vision. In all these domains, significant benefits including transparency and simplicity can be achieved by considering the physics of image formation and basic image science concepts 53 .…”

Section: Discussionmentioning

confidence: 99%

On the correlation between second order texture features and human observer detection performance in digital images

Nisbett

Kavuri

Das

2020

Sci Rep

View full text Add to dashboard Cite

image texture, the relative spatial arrangement of intensity values in an image, encodes valuable information about the scene. As it stands, much of this potential information remains untapped. Understanding how to decipher textural details would afford another method of extracting knowledge of the physical world from images. In this work, we attempt to bridge the gap in research between quantitative texture analysis and the visual perception of textures. the impact of changes in image texture on human observer's ability to perform signal detection and localization tasks in complex digital images is not understood. We examine this critical question by studying task-based human observer performance in detecting and localizing signals in tomographic breast images. We have also investigated how these changes impact the formation of second-order image texture. We used digital breast tomosynthesis (DBT) an FDA approved tomographic X-ray breast imaging method as the modality of choice to show our preliminary results. our human observer studies involve localization ROC (LROC) studies for low contrast mass detection in DBT. Simulated images are used as they offer the benefit of known ground truth. Our results prove that changes in system geometry or processing leads to changes in image texture magnitudes. We show that the variations in several well-known texture features estimated in digital images correlate with human observer detection-localization performance for signals embedded in them. This insight can allow efficient and practical techniques to identify the best imaging system design and algorithms or filtering tools by examining the changes in these texture features. This concept linking texture feature estimates and task based image quality assessment can be extended to several other imaging modalities and applications as well. it can also offer feedback in system and algorithm designs with a goal to improve perceptual benefits. Broader impact can be in wide array of areas including imaging system design, image processing, data science, machine learning, computer vision, perceptual and vision science. our results also point to the caution that must be exercised in using these texture features as image-based radiomic features or as predictive markers for risk assessment as they are sensitive to system or image processing changes.

show abstract

Section: Discussionmentioning

confidence: 99%

On the correlation between second order texture features and human observer detection performance in digital images

Nisbett

Kavuri

Das

2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…These parameters were also not present in the DICOM meta-data. Modern CT scanners often use varying strengths of IR to suppress image noise, but with the application of IR, images appear smoother, and depending on the strength, the noise texture becomes finer or more coarse (56)(57)(58). Another source of bias is the quality of the annotations and accuracy of information released about a public dataset (59).…”

mentioning

confidence: 99%

Detecting Spurious Correlations With Sanity Tests for Artificial Intelligence Guided Radiology Systems

Mahmood

Shrestha

Bates

et al. 2021

Front. Digit. Health

View full text Add to dashboard Cite

Artificial intelligence (AI) has been successful at solving numerous problems in machine perception. In radiology, AI systems are rapidly evolving and show progress in guiding treatment decisions, diagnosing, localizing disease on medical images, and improving radiologists' efficiency. A critical component to deploying AI in radiology is to gain confidence in a developed system's efficacy and safety. The current gold standard approach is to conduct an analytical validation of performance on a generalization dataset from one or more institutions, followed by a clinical validation study of the system's efficacy during deployment. Clinical validation studies are time-consuming, and best practices dictate limited re-use of analytical validation data, so it is ideal to know ahead of time if a system is likely to fail analytical or clinical validation. In this paper, we describe a series of sanity tests to identify when a system performs well on development data for the wrong reasons. We illustrate the sanity tests' value by designing a deep learning system to classify pancreatic cancer seen in computed tomography scans.

show abstract

“…Barrett reviews several basic concepts from image science that, from the author’s perspective, may be useful in designing and validating AI-based imaging systems. 2 The state-of-the-art deep-learning-based AI is commonly viewed as a “black box,” which has been empirically found to provide useful solutions to many practical problems, but the scientific mechanisms behind these solutions are not well understood. The theoretical perspectives provided in Barrett’s article encourage further research in understanding AI in connection with image science wisdom.…”

mentioning

confidence: 99%

Special Section Guest Editorial: Evaluation Methodologies for Clinical AI

et al. 2020

View full text Add to dashboard Cite

With the explosion of deep learning applications in medical imaging there is an urgent need to develop methods to evaluate the performance of artificial intelligence (AI) systems due to the increased complexities/varieties of AI technologies, the dependence of these new technologies on large datasets, and the emergence of novel types of clinical applications of AI systems. Proper testing methodology, metrics, appropriate training/tuning/validation study designs, and statistical analysis methods are needed to ensure that studies produce meaningful, robust, and generalizable results in a least burdensome fashion. These elements are key to the clinical adoption of AI technologies. Thus this Special Section for the Journal of Medical Imaging, Volume 7, Issue 1, encouraged relevant submissions in these topic areas. AI is not new to medical imaging. Since the earliest days of the SPIE Medical Imaging symposium there have been presentations on what was then referred to as Computer-Aided Diagnosis (CAD). The Computer-Aided Diagnosis conference at the larger SPIE Medical Imaging (MI) symposium was launched in 2006. Applications for CAD in mammography, lung CT, and chest x-ray imaging, all mature commercial products today, were discussed in their earliest phases at this conference. SPIE MI has also been the home for the introduction of new approaches to the assessment of CAD algorithms, a tradition that continues primarily through the conference on Image Perception, Observer Performance, and Technology Assessment. A perusal of the SPIE MI program through the years allows the reader to see the progression of AI algorithm development as well as methods for AI assessment. What is new to AI is the recent advance in computational power and the availability of large datasets that have enabled the successful application of deep neural network (DNN) architectures for various medical imaging tasks. These tasks include the common applications in the field related to the finding of suspicious areas in images for a reader to give a second-look, as well as the characterization of a reader's identified areas of suspicion with the support of AI. Newer tasks to which DNNs are being applied include image denoising, full image reconstruction from highly sparse or very noisy projections, triage systems that alert the user to high-priority cases so as to adjust casereading order, AI-selected image acquisition parameters on a per-patient basis, and the approximation of the ideal observer for use as a measure of image quality in complex imaging scenarios. For some applications, the performance of AI is being demonstrated to reach or surpass expert human performance such that automated diagnosis in which the clinician is replaced by the AI system is arguably close at hand. Moreover, the range of imaging modalities for which AI is being applied is vast, from the x-ray applications listed above, to optical, ultrasound, MRI, and digital pathology, the latter recently introduced as a conference track of its own at the SPIE MI symposium. Across the w...

show abstract

Is there a role for image science in the brave new world of artificial intelligence?

Abstract: Is there a role for image science in the brave new world of artificial intelligence?,"Abstract. Knowledge of the principles of image science is essential to the successful application of artificial intelligence in medical imaging.

Cited by 5 publications

References 33 publications

On the correlation between second order texture features and human observer detection performance in digital images

On the correlation between second order texture features and human observer detection performance in digital images

Detecting Spurious Correlations With Sanity Tests for Artificial Intelligence Guided Radiology Systems

Special Section Guest Editorial: Evaluation Methodologies for Clinical AI

Contact Info

Product

Resources

About