Artificial Intelligence in Lymphoma PET Imaging

Hasani, Navid; Paravastu, Sriram; Farhadi, Faraz; Yousefirizi, Fereshteh; Morris, Michael; Rahmim, Arman; Roschewski, Mark; Summers, Ronald M.; Saboury, Babak

doi:10.1016/j.cpet.2021.09.006

Cited by 24 publications

(10 citation statements)

References 103 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Use of semi‐automated work‐flow models using both ML and human/physician interventions (e.g. physician detection of lesions followed by the use of algorithms to segment them) may be useful to improve accuracy 42 . Standardized processes and regulations are warranted for more widespread adoption, and the future of ML should be focused on prospective investigation.…”

Section: Discussionmentioning

confidence: 99%

“…40 A comprehensive review of the literature in this area is described elsewhere. 41,42 Here, we will briefly discuss the examples of how machine learning algorithms were applied to automate the measurement of tMTV. Capobianco et al 43 measured pretreatment tMTV of 301 patients with DLBCL treated on the REMARC trial, 44 using a CNN-based algorithm and evaluated if this automated calculation of tMTV correlates with tMTV measured manually by two experienced nuclear medicine physicians.…”

Section: Achi N E L E a R N I Ng I N Ly M Phom A R A Diologymentioning

confidence: 99%

See 1 more Smart Citation

Clinical approaches for integrating machine learning for patients with lymphoma: Current strategies and future perspectives

2023

View full text Add to dashboard Cite

With the advent of electronic medical records and the expansion of health care data, researchers, clinicians and patients, are now facing an unprecedented wealth of digital information. In fact, the complexity and volume of the data from basic science to clinical information are beyond the capacity of a human's brain to analyse and comprehend. 1 Furthermore, clinical management from diagnosis through staging and the consideration of treatment options is also becoming more challenging for patients and providers; hence we often reference guidelines informed by disease experts to serve patients. 2 Machine learning (ML) technology can offer tools to support the processing of large volumes of data and can potentially help us in various areas including supporting diagnosis, creating prognostic models and informing treatment decisions.The term 'machine learning' was first used in 1959 by Arthur Samuel when he programmed a digital computer to learn to play a better game of checkers. 3 ML is a subdomain of artificial intelligence and includes deep learning. This technology has been applied extensively in almost every facet of life including marketing, online advertising, global positioning satellite software, fraud detection, engineering, and is now becoming a major research field and method in medicine. Although the vision of computation augmenting or replacing the intellectual function of the physicians was discussed in the 1970s, 4 there is currently little in health care that has been driven by ML. However, research papers using ML algorithms have been exploding, particularly in the last 3 years. With the expansion of databases using electronic medical records, we are in the early days of applying ML methods to a greater extent in medicine. Overviews of ML in medicine or in oncology have been described elsewhere. 2,5,6 In this review, we will focus on summarizing how ML algorithms have been applied in various fields of lymphoma clinical research and discuss limitations and future directions.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Achi N E L E a R N I Ng I N Ly M Phom A R A Diologymentioning

confidence: 99%

Clinical approaches for integrating machine learning for patients with lymphoma: Current strategies and future perspectives

2023

View full text Add to dashboard Cite

show abstract

“…Opportunities for clinical use of AI in nuclear medicine practice were extensively reviewed recently, including brain imaging ( 45 ), head and neck imaging ( 46 ), lung imaging ( 47 ), cardiac imaging ( 48 , 49 ), vascular imaging ( 49 , 50 ), bone imaging ( 51 ), prostate imaging ( 52 ), and imaging of lymphoma ( 53 ). Neuroendocrine tumors, other cancers (including gastrointestinal, pancreatic, hepatobiliary, sarcoma, and hereditary), infection, and inflammation are some examples of additional areas requiring further consideration.…”

Section: Opportunitiesmentioning

confidence: 99%

Artificial Intelligence in Nuclear Medicine: Opportunities, Challenges, and Responsibilities Toward a Trustworthy Ecosystem

Saboury

Bradshaw

Boellaard³

et al. 2022

J Nucl Med

Self Cite

View full text Add to dashboard Cite

show abstract

“…We considered different supervision levels by changing the effect of supervised term of loss function for training. Lymphoma lesions segmentation from the whole-body FDG PET images is challenging since the number, size, site and shape of lesions are varied in patients 8,9 .…”

Section: Introductionmentioning

confidence: 99%

Revisiting the supervision level in semi-supervised learning for automated tumor segmentation: application to lymphoma FDG PET imaging

Yousefirizi¹,

Bloise³

et al. 2023

Medical Imaging 2023: Image Processing

Self Cite

View full text Add to dashboard Cite

Convolutional neural networks have shown the power to segment FDG-avid tumors from PET scans. However, the training step of most of these approaches is supervised, demanding annotated data for training. To reduce this need, we previously evaluated the performance of 3D U-Net for lesion segmentation by two semi-supervised approaches using the combined loss functions of Robust FCM (RFCM) and Mumford-Shah (MS) losses for unsupervised term, and Dice and Labeled FCM (LFCM) for supervised term. In this work, we applied two combined losses (RFCM+αLFCM and MS+αDice) to consider the effect of supervision level (α) on segmentation performance on a multi-center (BC&SM) dataset. We applied two experiments utilizing the PET images of 292 patients. In experiment I, training and test data are from diffuse large B-cell (DLBCL) cases mostly from center BC. In experiment II, we applied a more realistic scenario in which the segmentation model was trained on DLBCL and primary mediastinal B-cell (PMBCL) data mostly from BC and tested on SM data. Our results showed that increasing the impact of the supervised term leads to different segmentation performances in both experiments. The best Dice score for MS-based semi-supervised approach were 0.60±0.17 (α=0.7) and 0.6±0.08 (α=0.2) for experiments I and II respectively. In FCM-based approach the best Dice score were 0.71±0.09 (α=0.7) and 0.69±0.03 (α=0.3) for experiments I and II, respectively. We concluded when test data from a center has low contribution in training data, increasing the supervision level results in lower segmentation performance. The performance drops higher by increasing α in MS-based semi-supervised approach.

show abstract

Artificial Intelligence in Lymphoma PET Imaging

Cited by 24 publications

References 103 publications

Clinical approaches for integrating machine learning for patients with lymphoma: Current strategies and future perspectives

Clinical approaches for integrating machine learning for patients with lymphoma: Current strategies and future perspectives

Artificial Intelligence in Nuclear Medicine: Opportunities, Challenges, and Responsibilities Toward a Trustworthy Ecosystem

Revisiting the supervision level in semi-supervised learning for automated tumor segmentation: application to lymphoma FDG PET imaging

Contact Info

Product

Resources

About