1
|
Mikulić M, Vičević D, Nagy E, Napravnik M, Štajduhar I, Tschauner S, Hržić F. Balancing Performance and Interpretability in Medical Image Analysis: Case study of Osteopenia. JOURNAL OF IMAGING INFORMATICS IN MEDICINE 2024:10.1007/s10278-024-01194-8. [PMID: 39020155 DOI: 10.1007/s10278-024-01194-8] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Received: 03/26/2024] [Revised: 06/09/2024] [Accepted: 06/25/2024] [Indexed: 07/19/2024]
Abstract
Multiple studies within the medical field have highlighted the remarkable effectiveness of using convolutional neural networks for predicting medical conditions, sometimes even surpassing that of medical professionals. Despite their great performance, convolutional neural networks operate as black boxes, potentially arriving at correct conclusions for incorrect reasons or areas of focus. Our work explores the possibility of mitigating this phenomenon by identifying and occluding confounding variables within images. Specifically, we focused on the prediction of osteopenia, a serious medical condition, using the publicly available GRAZPEDWRI-DX dataset. After detection of the confounding variables in the dataset, we generated masks that occlude regions of images associated with those variables. By doing so, models were forced to focus on different parts of the images for classification. Model evaluation using F1-score, precision, and recall showed that models trained on non-occluded images typically outperformed models trained on occluded images. However, a test where radiologists had to choose a model based on the focused regions extracted by the GRAD-CAM method showcased different outcomes. The radiologists' preference shifted towards models trained on the occluded images. These results suggest that while occluding confounding variables may degrade model performance, it enhances interpretability, providing more reliable insights into the reasoning behind predictions. The code to repeat our experiment is available on the following link: https://github.com/mikulicmateo/osteopenia .
Collapse
Affiliation(s)
- Mateo Mikulić
- University of Rijeka, Faculty of Engineering, Department of Computer Engineering, Vukovarska 58, Rijeka, 51000, Croatia
| | - Dominik Vičević
- University of Rijeka, Faculty of Engineering, Department of Computer Engineering, Vukovarska 58, Rijeka, 51000, Croatia
| | - Eszter Nagy
- Medical University of Graz, Department of Radiology, Division of Pediatric Radiology, Graz, 8036, Austria
| | - Mateja Napravnik
- University of Rijeka, Faculty of Engineering, Department of Computer Engineering, Vukovarska 58, Rijeka, 51000, Croatia
| | - Ivan Štajduhar
- University of Rijeka, Faculty of Engineering, Department of Computer Engineering, Vukovarska 58, Rijeka, 51000, Croatia
- University of Rijeka, Center for Artificial Intelligence and Cybersecurity, Radmile Matejčić 2, Rijeka, 51000, Croatia
| | - Sebastian Tschauner
- Medical University of Graz, Department of Radiology, Division of Pediatric Radiology, Graz, 8036, Austria
| | - Franko Hržić
- University of Rijeka, Faculty of Engineering, Department of Computer Engineering, Vukovarska 58, Rijeka, 51000, Croatia.
- University of Rijeka, Center for Artificial Intelligence and Cybersecurity, Radmile Matejčić 2, Rijeka, 51000, Croatia.
| |
Collapse
|
2
|
Birdal O, İpek E, Saygı M, Doğan R, Pay L, Tanboğa IH. Cluster analysis of clinical, angiographic, and laboratory parameters in patients with ST-segment elevation myocardial infarction. Lipids Health Dis 2024; 23:166. [PMID: 38835073 DOI: 10.1186/s12944-024-02128-7] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/07/2024] [Accepted: 04/30/2024] [Indexed: 06/06/2024] Open
Abstract
INTRODUCTION ST-segment elevation myocardial infarction (STEMI) represents the most harmful clinical manifestation of coronary artery disease. Risk assessment plays a beneficial role in determining both the treatment approach and the appropriate time for discharge. Hierarchical agglomerative clustering (HAC), a machine learning algorithm, is an innovative approach employed for the categorization of patients with comparable clinical and laboratory features. The aim of the present study was to investigate the role of HAC in categorizing STEMI patients and to compare the results of these patients. METHODS A total of 3205 patients who were diagnosed with STEMI at the university hospital emergency clinic between 2015 and 2023 were included in the study. The patients were divided into 2 different phenotypic disease clusters using the HAC method, and their outcomes were compared. RESULTS In the present study, a total of 3205 STEMI patients were included; 2731 patients were in cluster 1, and 474 patients were in cluster 2. Mortality was observed in 147 (5.4%) patients in cluster 1 and 108 (23%) patients in cluster 2 (chi-square P value < 0.01). Survival analysis revealed that patients in cluster 2 had a significantly greater risk of death than patients in cluster 1 did (log-rank P < 0.001). After adjustment for age and sex in the Cox proportional hazards model, cluster 2 exhibited a notably greater risk of death than did cluster 1 (HR = 3.51, 95% CI = 2.71-4.54; P < 0.001). CONCLUSION Our study showed that the HAC method may be a potential tool for predicting one-month mortality in STEMI patients.
Collapse
Affiliation(s)
- Oğuzhan Birdal
- Department of Cardiology, Ataturk University, Erzurum, 25240, Turkey.
| | - Emrah İpek
- Department of First Aid and Emergency, Health Services Vocational School, Nisantasi University, Istanbul, 34360, Turkey
| | - Mehmet Saygı
- Department of Cardiology, Hisar Intercontinental Hospital, Istanbul, 34764, Turkey
| | - Remziye Doğan
- Department of Cardiology, Hisar Intercontinental Hospital, Istanbul, 34764, Turkey
| | - Levent Pay
- Department of Cardiology, Ardahan State Hospital, Ardahan, 75000, Turkey
| | - Ibrahim Halil Tanboğa
- Department of Cardiology and Biostatistics, Nisantasi University, Istanbul, 34360, Turkey
| |
Collapse
|
3
|
Wu H, Chen Z, Gu J, Jiang Y, Gao S, Chen W, Miao C. Predicting Chronic Pain and Treatment Outcomes Using Machine Learning Models Based on High-dimensional Clinical Data From a Large Retrospective Cohort. Clin Ther 2024; 46:490-498. [PMID: 38824080 DOI: 10.1016/j.clinthera.2024.04.012] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/29/2023] [Revised: 04/13/2024] [Accepted: 04/27/2024] [Indexed: 06/03/2024]
Abstract
PURPOSE To identify factors and indicators that affect chronic pain and pain relief, and to develop predictive models using machine learning. METHODS We analyzed the data of 67,028 outpatient cases and 11,310 valid samples with pain from a large retrospective cohort. We used decision tree, random forest, AdaBoost, neural network, and logistic regression to discover significant indicators and to predict pain and treatment relief. FINDINGS The random forest model had the highest accuracy, F1 value, precision, and recall rates for predicting pain relief. The main factors affecting pain and treatment relief included body mass index, blood pressure, age, body temperature, heart rate, pulse, and neutrophil/lymphocyte × platelet ratio. The logistic regression model had high sensitivity and specificity for predicting pain occurrence. IMPLICATIONS Machine learning models can be used to analyze the risk factors and predictors of chronic pain and pain relief, and to provide personalized and evidence-based pain management.
Collapse
Affiliation(s)
- Han Wu
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China
| | - Zhaoyuan Chen
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China
| | - Jiahui Gu
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China
| | - Yi Jiang
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China
| | - Shenjia Gao
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China
| | - Wankun Chen
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China.
| | - Changhong Miao
- Department of Anesthesiology, Zhongshan Hospital, Fudan University, Shanghai, China; Shanghai Key laboratory of Perioperative Stress and Protection, Shanghai, China.
| |
Collapse
|
4
|
Evensen S, Taraldsen K, Aam S, Morandi A. Delirium is associated with low levels of upright activity in geriatric inpatients-results from a prospective observational study. Aging Clin Exp Res 2024; 36:41. [PMID: 38353776 PMCID: PMC10867047 DOI: 10.1007/s40520-024-02699-6] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/27/2023] [Accepted: 01/08/2024] [Indexed: 02/16/2024]
Abstract
BACKGROUND Delirium is common in geriatric inpatients and associated with poor outcomes. Hospitalization is associated with low levels of physical activity. Motor symptoms are common in delirium, but how delirium affects physical activity remains unknown. AIMS To investigate differences in physical activity between geriatric inpatients with and without delirium. METHODS We included acutely admitted patients ≥ 75 years in a prospective observational study at a medical geriatric ward at a Norwegian University Hospital. Delirium was diagnosed according to the DSM-5 criteria. Physical activity was measured by an accelerometer-based device worn on the right thigh. The main outcome was time in upright position (upright time) per 24 h (00.00 to 23.59) on the first day of hospitalization with verified delirium status. Group differences were analysed using t test. RESULTS We included 237 patients, mean age 86.1 years (Standard Deviation (SD) 5.1), and 73 patients (30.8%) had delirium. Mean upright time day 1 for the entire group was 92.2 min (SD 84.3), with 50.9 min (SD 50.7) in the delirium group and 110.6 min (SD 89.7) in the no-delirium group, mean difference 59.7 minutes, 95% Confidence Interval 41.6 to 77.8, p value < 0.001. DISCUSSION Low levels of physical activity in patients with delirium raise the question if immobilization may contribute to poor outcomes in delirium. Future studies should investigate if mobilization interventions could improve outcomes of delirium. CONCLUSIONS In this sample of geriatric inpatients, the group with delirium had lower levels of physical activity than the group without delirium.
Collapse
Affiliation(s)
- Sigurd Evensen
- Department of Medicine, Diakonhjemmet Hospital, Oslo, Norway.
| | - Kristin Taraldsen
- Department of Rehabilitation Science and Health Technology, Oslo Metropolitan University (OsloMet), Oslo, Norway
| | - Stina Aam
- Department of Geriatric Medicine, Clinic of Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway
- Department of Neuromedicine and Movement Science, Faculty of Medicine and Health Service, NTNU-Norwegian University of Science and Technology, Trondheim, Norway
| | - Alessandro Morandi
- Intermediate Care and Rehabilitation, Azienda Speciale Di Cremona Solidale, Cremona Parc Sanitari Pere Virgili, Cremona, Italy
- Vall d'Hebrón Institute of Research, Barcelona, Spain
| |
Collapse
|
5
|
Woodman RJ, Koczwara B, Mangoni AA. Applying precision medicine principles to the management of multimorbidity: the utility of comorbidity networks, graph machine learning, and knowledge graphs. Front Med (Lausanne) 2024; 10:1302844. [PMID: 38404463 PMCID: PMC10885565 DOI: 10.3389/fmed.2023.1302844] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/27/2023] [Accepted: 12/22/2023] [Indexed: 02/27/2024] Open
Abstract
The current management of patients with multimorbidity is suboptimal, with either a single-disease approach to care or treatment guideline adaptations that result in poor adherence due to their complexity. Although this has resulted in calls for more holistic and personalized approaches to prescribing, progress toward these goals has remained slow. With the rapid advancement of machine learning (ML) methods, promising approaches now also exist to accelerate the advance of precision medicine in multimorbidity. These include analyzing disease comorbidity networks, using knowledge graphs that integrate knowledge from different medical domains, and applying network analysis and graph ML. Multimorbidity disease networks have been used to improve disease diagnosis, treatment recommendations, and patient prognosis. Knowledge graphs that combine different medical entities connected by multiple relationship types integrate data from different sources, allowing for complex interactions and creating a continuous flow of information. Network analysis and graph ML can then extract the topology and structure of networks and reveal hidden properties, including disease phenotypes, network hubs, and pathways; predict drugs for repurposing; and determine safe and more holistic treatments. In this article, we describe the basic concepts of creating bipartite and unipartite disease and patient networks and review the use of knowledge graphs, graph algorithms, graph embedding methods, and graph ML within the context of multimorbidity. Specifically, we provide an overview of the application of graph theory for studying multimorbidity, the methods employed to extract knowledge from graphs, and examples of the application of disease networks for determining the structure and pathways of multimorbidity, identifying disease phenotypes, predicting health outcomes, and selecting safe and effective treatments. In today's modern data-hungry, ML-focused world, such network-based techniques are likely to be at the forefront of developing robust clinical decision support tools for safer and more holistic approaches to treating older patients with multimorbidity.
Collapse
Affiliation(s)
- Richard John Woodman
- Flinders Health and Medical Research Institute, College of Medicine and Public Health, Flinders University, Adelaide, SA, Australia
| | - Bogda Koczwara
- Flinders Health and Medical Research Institute, College of Medicine and Public Health, Flinders University, Adelaide, SA, Australia
- Department of Medical Oncology, Flinders Medical Centre, Southern Adelaide Local Health Network, Adelaide, SA, Australia
| | - Arduino Aleksander Mangoni
- Flinders Health and Medical Research Institute, College of Medicine and Public Health, Flinders University, Adelaide, SA, Australia
- Department of Clinical Pharmacology, Flinders Medical Centre, Southern Adelaide Local Health Network, Adelaide, SA, Australia
| |
Collapse
|
6
|
Mondello A, Dal Bo M, Toffoli G, Polano M. Machine learning in onco-pharmacogenomics: a path to precision medicine with many challenges. Front Pharmacol 2024; 14:1260276. [PMID: 38264526 PMCID: PMC10803549 DOI: 10.3389/fphar.2023.1260276] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/17/2023] [Accepted: 12/26/2023] [Indexed: 01/25/2024] Open
Abstract
Over the past two decades, Next-Generation Sequencing (NGS) has revolutionized the approach to cancer research. Applications of NGS include the identification of tumor specific alterations that can influence tumor pathobiology and also impact diagnosis, prognosis and therapeutic options. Pharmacogenomics (PGx) studies the role of inheritance of individual genetic patterns in drug response and has taken advantage of NGS technology as it provides access to high-throughput data that can, however, be difficult to manage. Machine learning (ML) has recently been used in the life sciences to discover hidden patterns from complex NGS data and to solve various PGx problems. In this review, we provide a comprehensive overview of the NGS approaches that can be employed and the different PGx studies implicating the use of NGS data. We also provide an excursus of the ML algorithms that can exert a role as fundamental strategies in the PGx field to improve personalized medicine in cancer.
Collapse
Affiliation(s)
| | | | | | - Maurizio Polano
- Experimental and Clinical Pharmacology Unit, Centro di Riferimento Oncologico di Aviano (CRO), Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS), Aviano, Italy
| |
Collapse
|