A human-interpretable machine learning pipeline based on ultrasound to support leiomyosarcoma diagnosis.

The preoperative evaluation of myometrial tumors is essential to avoid delayed treatment and to establish the appropriate surgical approach. Specifically, the differential diagnosis of leiomyosarcoma (LMS) is particularly challenging due to the overlapping of clinical, laboratory and ultrasound features between fibroids and LMS. In this work, we present a human-interpretable machine learning (ML) pipeline to support the preoperative differential diagnosis of LMS from leiomyomas, based on both clinical data and gynecological ultrasound assessment of 68 patients (8 with LMS diagnosis). The pipeline provides the following novel contributions: (i) end-users have been involved both in the definition of the ML tasks and in the evaluation of the overall approach; (ii) clinical specialists get a full understanding of both the decision-making mechanisms of the ML algorithms and the impact of the features on each automatic decision. Moreover, the proposed pipeline addresses some of the problems concerning both the imbalance of the two classes by analyzing and selecting the best combination of the synthetic oversampling strategy of the minority class and the classification algorithm among different choices, and the explainability of the features at global and local levels. The results show very high performance of the best strategy (AUC = 0.99, F1 = 0.87) and the strong and stable impact of two ultrasound-based features (i.e., tumor borders and consistency of the lesions). Furthermore, the SHAP algorithm was exploited to quantify the impact of the features at the local level and a specific module was developed to provide a template-based natural language (NL) translation of the explanations for enhancing their interpretability and fostering the use of ML in the clinical setting.

Corrigendum: Dietary customs and social deprivation in an aging population from Southern Italy: A machine learning approach.

[This corrects the article DOI: 10.3389/fnut.2022.811076.].

Dietary profiling of physical frailty in older age phenotypes using a machine learning approach: the Salus in Apulia Study.

PURPOSE: Growing awareness of the biological and clinical value of nutrition in frailty settings calls for further efforts to investigate dietary gaps to act sooner to achieve focused management of aging populations. We cross-sectionally examined the eating habits of an older Mediterranean population to profile dietary features most associated with physical frailty. METHODS: Clinical and physical examination, routine biomarkers, medical history, and anthropometry were analyzed in 1502 older adults (65 +). CHS criteria were applied to classify physical frailty, and a validated Food Frequency Questionnaire to assess diet. The population was subdivided by physical frailty status (frail or non-frail). Raw and adjusted logistic regression models were applied to three clusters of dietary variables (food groups, macronutrients, and micronutrients), previously selected by a LASSO approach to better predict diet-related frailty determinants. RESULTS: A lower consumption of wine (OR 0.998, 95% CI 0.997-0.999) and coffee (OR 0.994, 95% CI 0.989-0.999), as well as a cluster of macro and micronutrients led by PUFAs (OR 0.939, 95% CI 0.896-0.991), zinc (OR 0.977, 95% CI 0.952-0.998), and coumarins (OR 0.631, 95% CI 0.431-0.971), was predictive of non-frailty, but higher legumes intake (OR 1.005, 95%CI 1.000-1.009) of physical frailty, regardless of age, gender, and education level. CONCLUSIONS: Higher consumption of coffee and wine, as well as PUFAs, zinc, and coumarins, as opposed to legumes, may work well in protecting against a physical frailty profile of aging in a Mediterranean setting. Longitudinal investigations are needed to better understand the causal potential of diet as a modifiable contributor to frailty during aging.

Post-hoc analysis of a tool to predict kidney failure in patients with IgA nephropathy.

BACKGROUND: Recently, a tool based on two different artificial neural networks has been developed. The first network predicts kidney failure (KF) development while the second predicts the time frame to reach this outcome. In this study, we conducted a post-hoc analysis to evaluate the discordant results obtained by the tool. METHODS: The tool performance was analyzed in a retrospective cohort of 1116 adult IgAN patients, as were the causes of discordance between the predicted and observed cases of KF. RESULTS: There was discordance between the predicted and observed KF in 216 IgAN patients (19.35%) all of whom were elderly, hypertensive, had high serum creatinine levels, reduced renal function and moderate or severe renal lesions. Many of these patients did not receive therapy or were non-responders to therapy. In other IgAN patients the tool predicted KF but the outcome was not reached because patients responded to therapy. Therefore, in the discordant group (prediction did not match the observed outcome) the proportion of patients having or not having KF was strongly associated with treatment (P < 0.0001). CONCLUSIONS: The post-hoc analysis shows that discordance in a low number of patients is not an error, but rather the effect of positive response to therapy. Thus, the tool could both help physicians to determine the prognosis of the disease and help patients to plan for their future.

Prediction of chronic kidney disease and its progression by artificial intelligence algorithms.

BACKGROUND AND OBJECTIVE: Aim of nephrologists is to delay the outcome and reduce the number of patients undergoing renal failure (RF) by applying prevention protocols and accurately monitoring chronic kidney disease (CKD) patients. General practitioners and nephrologists are involved in the first and in the late stages of the disease, respectively. Early diagnosis of CKD is an important step in preventing the progression of kidney damage. Our aim was to review publications on machine learning algorithms (MLAs) that can predict early CKD and its progression. METHODS: We conducted a systematic review and selected 55 articles on the application of MLAs in CKD. PubMed, Medline, Scopus, Web of Science and IEEE Xplore Digital Library of the Institute of Electrical and Electronics Engineers were searched. The search terms were chronic kidney disease, artificial intelligence, data mining and machine learning algorithms. RESULTS: MLAs use enormous numbers of predictors combining them in non-linear and highly interactive ways. This ability increases when new data is added. We observed some limitations in the publications: (i) databases were not accurately reviewed by physicians; (ii) databases did not report the ethnicity of the patients; (iii) some databases collected variables that were not important for the diagnosis and progression of CKD; (iv) no information was presented on the native kidney disease causing CKD; (v) no validation of the results in external independent cohorts was provided; and (vi) no insights were given on the MLAs that were used. Overall, there was limited collaboration among experts in electronics, computer science and physicians. CONCLUSIONS: The application of MLAs in kidney diseases may enhance the ability of clinicians to predict CKD and RF, thus improving diagnostic assistance and providing suitable therapeutic decisions. However, it is necessary to improve the development process of MLA tools.

Dietary Patterns Associated with Diabetes in an Older Population from Southern Italy Using an Unsupervised Learning Approach.

Dietary behaviour is a core element in diabetes self-management. There are no remarkable differences between nutritional guidelines for people with type 2 diabetes and healthy eating recommendations for the general public. This study aimed to evaluate dietary differences between subjects with and without diabetes and to describe any emerging dietary patterns characterizing diabetic subjects. In this cross-sectional study conducted on older adults from Southern Italy, eating habits in the "Diabetic" and "Not Diabetic" groups were assessed with FFQ, and dietary patterns were derived using an unsupervised learning algorithm: principal component analysis. Diabetic subjects (n = 187) were more likely to be male, slightly older, and with a slightly lower level of education than subjects without diabetes. The diet of diabetic subjects reflected a high-frequency intake of dairy products, eggs, vegetables and greens, fresh fruit and nuts, and olive oil. On the other hand, the consumption of sweets and sugary foods was reduced compared to non-diabetics (23.74 ± 35.81 vs. 16.52 ± 22.87; 11.08 ± 21.85 vs. 7.22 ± 15.96). The subjects without diabetes had a higher consumption of red meat, processed meat, ready-to-eat dishes, alcoholic drinks, and lower vegetable consumption. The present study demonstrated that, in areas around the Mediterranean Sea, older subjects with diabetes had a healthier diet than their non-diabetic counterparts.

Dietary Customs and Social Deprivation in an Aging Population From Southern Italy: A Machine Learning Approach.

Background: Diet and social determinants influence the state of human health. In older adults, the presence of social, physical and psychological barriers increases the probability of deprivation. This study investigated the relationship between social deprivation and eating habits in non-institutionalized older adults from Southern Italy, and identified foods and dietary habits associated with social deprivation. Methods: We recruited 1,002 subjects, mean age 74 years, from the large population based Salus in Apulia Study. In this cross-sectional study, eating habits and the level of deprivation were assessed with FFQ and DiPCare-Q, respectively. Results: Deprived subjects (n = 441) included slightly more females, who were slightly older and with a lower level of education. They consumed less fish (23 vs. 26 g), fruiting vegetables (87 vs. 102 g), nuts (6 vs. 9 g) and less "ready to eat" dishes (29 vs. 33 g). A Random Forest (RF) model was used to identify a dietary pattern associated with social deprivation. This pattern included an increased consumption of low-fat dairy products and white meat, and a decreased consumption of wine, leafy vegetables, seafood/shellfish, processed meat, red meat, dairy products, and eggs. Conclusion: The present study showed that social factors also define diet and eating habits. Subjects with higher levels of deprivation consume cheaper and more readily available food.

Artificial intelligence in glomerular diseases.

In this narrative review, we focus on the application of artificial intelligence in the clinical history of patients with glomerular disease, digital pathology in kidney biopsy, renal ultrasonography imaging, and prediction of chronic kidney disease (CKD). With the development of natural language processing, the clinical history of a patient can be used to identify a computable phenotype. In kidney pathology, digital imaging has adopted innovative deep learning algorithms (DLAs) that can improve the predictive capability of the examined lesions. However, at this time, these applications can only be used in research because there is no recognized validation to replace the conventional diagnostic applications. Kidney ultrasonography, used in the clinical examination of patients, provides information about the progression of kidney damage. Machine learning algorithms (MLAs) with promising results for the early detection of CKD have been proposed, but, still, they are not solid enough to be incorporated into the clinical practice. A few tools for glomerulonephritis, based on MLAs, are available in clinical practice. They can be downloaded on computers and cellular phones but can only be applied to uniracial cohorts of patients. To improve their performance, it is necessary to organize large consortia with multiracial cohorts. Finally, in many studies MLA development has been carried out using retrospective cohorts. The performance of the models might differ in retrospective cohorts compared to real-world data. Therefore, the models should be validated in prospective external large cohorts.

Development and testing of an artificial intelligence tool for predicting end-stage kidney disease in patients with immunoglobulin A nephropathy.

We have developed an artificial neural network prediction model for end-stage kidney disease (ESKD) in patients with primary immunoglobulin A nephropathy (IgAN) using a retrospective cohort of 948 patients with IgAN. Our tool is based on a two-step procedure of a classifier model that predicts ESKD, and a regression model that predicts development of ESKD over time. The classifier model showed a performance value of 0.82 (area under the receiver operating characteristic curve) in patients with a follow-up of five years, which improved to 0.89 at the ten-year follow-up. Both models had a higher recall rate, which indicated the practicality of the tool. The regression model showed a mean absolute error of 1.78 years and a root mean square error of 2.15 years. Testing in an independent cohort of 167patients with IgAN found successful results for 91% of the patients. Comparison of our system with other mathematical models showed the highest discriminant Harrell C index at five- and ten-years follow-up (81% and 86%, respectively), paralleling the lowest Akaike information criterion values (355.01 and 269.56, respectively). Moreover, our system was the best calibrated model indicating that the predicted and observed outcome probabilities did not significantly differ. Finally, the dynamic discrimination indexes of our artificial neural network, expressed as the weighted average of time-dependent areas under the curve calculated at one and two years, were 0.80 and 0.79, respectively. Similar results were observed over a 25-year follow-up period. Thus, our tool identified individuals who were at a high risk of developing ESKD due to IgAN and predicted the time-to-event endpoint. Accurate prediction is an important step toward introduction of a therapeutic strategy for improving clinical outcomes.

A community-built calibration system: The case study of quantification of metabolites in grape juice by qNMR spectroscopy.

Nuclear Magnetic Resonance (NMR) is an analytical technique extensively used in almost every chemical laboratory for structural identification. This technique provides statistically equivalent signals in spite of using spectrometer with different hardware features and is successfully used for the traceability and quantification of analytes in food samples. Nevertheless, to date only a few internationally agreed guidelines have been reported on the use of NMR for quantitative analysis. The main goal of the present study is to provide a methodological pipeline to assess the reproducibility of NMR data produced for a given matrix by spectrometers from different manufacturers, with different magnetic field strengths, age and hardware configurations. The results have been analyzed through a sequence of chemometric tests to generate a community-built calibration system which was used to verify the performance of the spectrometers and the reproducibility of the predicted sample concentrations.

Clinical decision support system for end-stage kidney disease risk estimation in IgA nephropathy patients.

BACKGROUND: The progression of IgA nephropathy (IgAN) to end-stage kidney disease (ESKD) depends on several factors that are not quite clear and tangle the risk assessment. We aimed at developing a clinical decision support system (CDSS) for a quantitative risk assessment of ESKD and its timing using available clinical data at the time of renal biopsy. METHODS: We included a total of 1040 biopsy-proven IgAN patients with long-term follow-up from Italy (N = 546), Norway (N = 441) and Japan (N = 53). Of these, 241 patients reached ESKD: 104 Italian [median time to ESKD = 5 (3-9) years], 134 Norwegian [median time to ESKD = 6 (2-11) years] and 3 Japanese [median time to ESKD = 3 (2-12) years]. We independently trained and validated two cooperating artificial neural networks (ANNs) for predicting first the ESKD status and then the time to ESKD (defined as three categories: ≤ 3 years, between > 3 and 8 years and over 8 years). As inputs we used gender, age, histological grading, serum creatinine, 24-h proteinuria and hypertension at the time of renal biopsy. RESULTS: The ANNs demonstrated high performance for both the prediction of ESKD (with an AUC of 89.9, 93.3 and 100% in the Italian, Norwegian and Japanese IgAN population, respectively) and its timing (f-measure of 90.7% in the cohort from Italy and 70.8% in the one from Norway). We embedded the two ANNs in a CDSS available online (www.igan.net). Entering the clinical parameters at the time of renal biopsy, the CDSS returns as output the estimated risk and timing of ESKD for the patient. CONCLUSIONS: This CDSS provides useful additional information for identifying 'high-risk' IgAN patients and may help stratify them in the context of a personalized medicine approach.