Revolutionizing the Future of Urology Through the Use of AI and Machine Learning
Welcome to Health Connect, the podcast for health professionals where we will share the latest news and information on science and technology in the medical field. In this episode, we’ll talk about the use of Artificial Intelligence in the management and diagnosis of urologic cancers.
Did you know that BPH affects over 200 million men worldwide and it is the leading cause of lower urinary tract symptoms in aging males?1
Despite the availability of several questionnaires for the clinical prediction of benign prostatic hyperplasia, or BPH, current diagnostic methods lack reliability and accuracy, even when they consider multiple assessment parameters such as flow rates and symptom scores. Thus, there is a greater need to assess the severity of the disease more precisely and get better clinical outcomes before surgical intervention. This need has led researchers to develop intelligent systems that can assist in the screening, investigation, analysis of symptoms and signs, and eventual management of patients with BPH.1
In that regard, Artificial Intelligence, or AI, methodologies are more accurate in predicting and studying big data cohorts than traditional statistics. AI has aided evidence-based and personalized medical treatment by making a growing collection of patient data available to physicians. For example, with the help of predictive models that combine in a non-linear fashion, these systems have been able to differentiate BPH from prostate cancer.1
In a recent review and meta-analysis, Shah and collaborators sought to highlight key findings and features in four different studies that employed fuzzy, computer vision-based, and clinical data mining analysis systems. The study aimed to identify not only the initial and histopathological diagnosis of BPH, but also the level of severity of the disease and the recommended treatment.1
In this review, the study by Torshizi and colleagues employed fuzzy intelligent systems with two modules to diagnose the severity level of BPH and recommend appropriate therapies. In this case, while the first module determined the degree of severity, the second one suggested a treatment decision based on ontology modeling, which used the results of the first module and added external knowledge as well.1
To validate the efficiency and accuracy of the system, a case study was conducted on 44 participants; these results were then compared with the decisions and recommendations of an expert panel. The model achieved an accuracy of 90%, suggesting that the decisions made by both the fuzzy system and the expert panel matched in nine out of ten cases.1
Meanwhile, Khalid and colleagues evaluated the application of a computer vision-based system for the histopathological diagnosis of BPH using 59 digital microphotograph pictures divided into training, validation, and test dataset. In these images, 169 regions were marked as sites of hyperplasia of the prostatic glandular component; this particular system achieved 96.3% accuracy in identifying the hyperplastic glandular areas.1
Thus, these results showed that the accuracy of fuzzy systems to correctly diagnose BPH was 90%, while that of computer-based vision systems was 96.3%. Data mining, on the other hand, achieved sensitivity and specificity of 70 and 50 percent, respectively, in correctly predicting the clinical response to medical treatment in BPH.1
These studies, and their respective results, show that not only is Artificial Intelligence gaining traction in urology, but also has the potential to improve diagnostics and patient care. There’s an increased interest in utilizing AI in screening, symptom analysis, investigation, and prognosis of BPH, as well as other urologic pathologies1 like cancer, since prognosis assessment is essential to determine the risk profile of patients with urologic cancers.2
To that end, Eminaga and collaborators developed a risk profile reconstruction model for cancer-specific survival estimation for continuous-time points after a urologic cancer diagnosis. Using AI-based algorithms, a national cancer registry data, and accessible clinical parameters for the risk profile reconstruction, the team derived a risk stratification model to estimate the minimum follow-up duration and the likelihood of risk stability in prostate, kidney, and testicular cancers.2
This longitudinal risk profile for cancer-specific mortality was constructed based on clinicopathological information obtained around the time of diagnosis using the Surveillance, Epidemiology, and End-Results, or SEER, database. The goal was to identify patients diagnosed with genitourinary malignancies, which yielded approximately 1.9 million patients, allowing an assignment of 90% of patients to a development set, and the remaining 10% to a test set.2
The research gathered data about tumor type, extent, staging, grading, biomarkers, age at diagnosis, and time from diagnosis. Cancer-specific mortality status at a given time point was also obtained. Estimating the time series of true survival probabilities in the population as a reference for model calibration was feasible given the large cohort size of the development set, which was more than one million patients. Nevertheless, for all patients, the true endpoint was cancer-related death status. The model output was the prediction score for cancer-specific mortality.2
Out of all the cases obtained through the database, 14.3% died due to one of the urologic cancers, with an overall median follow-up of 12 years. When patients were stratified by age at diagnosis, the age group of 30 to 64-years old had a median follow-up of 14 years, while the age group of 65 to 69-year-olds had a median follow-up of 13 years.2
By stratifying according to the cancer-specific death status, however, the median time to cancer-specific death was 11 years, while the median observation time was 13 years for those who were event-free at the time. This study by Eminaga and colleagues introduces and validates the risk-profile reconstruction concept of a population using advanced machine learning to estimate the cancer-specific survival probability.2
This prognostic model can accurately estimate the risk for cancer-specific mortality and provides a discriminative risk stratification. Furthermore, this model was well-calibrated for cancer-specific survival estimation over a long period, while providing stability and good fitness. For instance, the data-driven solution suggested at least a four-year follow-up after diagnosis for prostate cancer or five years for kidney and testicular cancers. In the upper quartile of the whole simulation cohorts, follow-up durations can reach 11 years for prostate cancers, 8 years for kidney cancers, or 7 years for testicular cancers.2
Other possible follow-up durations can reach up to 10 years for prostate cancers in 75% of simulated cases, regardless of the tumor dissemination status. In kidney cancers, 75% simulated cases can have a maximum follow-up duration of 9 years, regardless of the tumor dissemination status. Finally, the simulation study revealed that follow-up can continue up to 8 years for localized testicular cancers or 9 years for testicular cancers with distant metastasis.2
CÁPSULA
Artificial intelligence was designed to mimic and perform human cognitive tasks, and it is shifting the way clinicians make decisions and the way we provide healthcare. It is a rapidly advancing field that is very likely to become an integral part of the healthcare system, thanks to AI's ability to process large amounts of data with high accuracy and provision of reliable cues for informed decision-making.1
Genitourinary malignancies are among the most common cancers, with a wide variety of presentations, risk profiles, and prognosis.2 In recent years, AI and machine learning have been increasingly studied in the field of urologic cancers in areas such as diagnosis, radiomics, risk stratification, tissue sample analysis also known as pathomics, surgical skill assessment, patient outcomes and prognosis, mortality and recurrence prediction, and clinical decision making, to name a few.3
FIN DE CÁPSULA
As we’ve seen so far, AI has the potential to revolutionize the future of cancer management by improving diagnostic accuracy, treatment planning, and patient outcomes, since it is feasible to develop a data-driven solution for cancer-specific survival estimation and the potential follow-up management of urologic cancers.2,3
In terms of risk stratification, some works seek to assess the likelihood of cancer progression and help determine the appropriate course of treatment. Depending on the risk level, it may include active surveillance, surgery, radiation therapy, and/or hormone therapy.3
To that end, Winkel and collaborators investigated whether machine learning algorithms in combination with biparametric imaging could accurately detect and classify prostate lesions in asymptomatic men. They trained a model using a cohort of 48 men, 38 of whom had high-risk lesions, while 10 were lesion-free. The model identified and classified 100% of the high-risk lesions and 73% of the intermediate-risk lesions.3
In the diagnosis of prostatic cancer, on the other hand, multiparametric prostate MRI has become a more widely used measurement of clinical significance. The Prostate Imaging Reporting and Data System, commonly known as PI-RADS, is an international standard for acquiring, interpreting, and reporting MRI images of the prostate.3
PI-RADS-lesion categories 1 and 2 usually indicate lower chances of clinically significant cancer, while PI-RADS 4 and 5 usually indicate the opposite. Category 3 lesions pose challenges to clinicians and radiologists alike due to the ambiguity of this designation in terms of clinically significant or insignificant prostate cancer. However, there’s growing evidence that machine learning models can rival the performance of clinical radiologists in assessing PI-RADS lesions by offering the means to clearly distinguish between clinically significant and insignificant prostate cancer.3
In this regard, Hectors and collaborators used machine learning to construct and cross-validate a model using radiomic features from T2-weighted imaging of PIRADS 3 lesions to identify clinically significant prostate cancer. Using a training set of 188 subjects and a test set of 52 subjects, they were able to train a random forest classifier with an area under the curve of 0.76 for predicting clinically significant prostate cancer in the test set.3
Multimodal machine learning approaches blend the strengths of different imaging modalities, like MRI or CT, to optimize the visualization of anatomic structures with other modalities that emphasize function, such as positron emission tomography or ultrasonography. For instance, in 2021, Khosravi and team utilized an AI-driven approach that combined MRI and histopathologic data from biopsy reports to increase the accuracy of PI-RADS scoring, while a different study employed multiparametric MRI as a prescreening test before a transrectal ultrasound-guided biopsy among men with clinical suspicion of prostate cancer.3
In terms of prognosis, AI has been evaluated for the prediction of survival and mortality, as well as recurrence. The ability of machine learning models to rapidly process large amounts of data allows for its increased prognostic capability in prostate cancer. A machine learning model developed by Bibault and colleagues consisted of 30 clinical features and predicted the risk of prostate cancer mortality within 10 years of diagnosis with an accuracy of 0.98. On this model’s prediction, Gleason score, PSA at diagnosis, and age had the largest impact on prediction.3
For survival modeling, on the other hand, several machine learning approaches have been introduced in the last few years, and most continue to respect the proportional hazards assumption or apply a predictive model at fixed time points. Random Survival Forests, DeepSurv, and DeepHit are examples of models that have been evaluated for survival prediction.2
In addition to mortality and survival prognosis, recurrence risk prediction needs to be investigated following radical prostatectomies. The recurrence in prostate cancer patients after a radical prostatectomy often couples with a higher mortality rate. Machine-learning algorithms have shown higher accuracy rates than predictive nomograms, previously used in predicting prostate cancer recurrence.3
An example of this was seen in the work by Tan and colleagues, where three machine learning models outperformed traditional nomograms in predicting biochemical recurrence at one, three, and five years. The best model reached an area under the curve of 0.894 for a five-year recurrence. This provides an alternative to tailored care in multimodal therapy.3
AI has also been used in the informed decision-making process. Auffenberg and collaborators developed a web-based system allowing patients to input their specific information to generate treatment options. This system was trained using a random forest model processing data from more than 7,000 prostate cancer patients covering different therapies, including radical prostatectomy and radiation therapy. Both patients and doctors could benefit from employing this tool to help them make informed decisions about the best-tailored treatment for each patient.3
A different machine learning approach, artificial neural network, has shown adequate predictive performance and can grasp the inherent data patterns more effectively than traditional statistical methods. This model has been applied widely in urology for the differentiation between tumor grade or subtype of genitourinary malignancies, the prediction of treatment response and tumor recurrence, as well as patient survival.4
In the clinical practice of urolithiasis, the most common use of artificial neural network models lies in the prediction of endourologic surgical outcomes and stone-free status after Extracorporeal Shock Wave Lithotripsy.4
In this regard, Hong and colleagues undertook the construction of a urosepsis risk prediction model based on ultrasound and urinalysis for patients with upper urinary tract calculi, using the artificial neural network data mining approach. The objective was to adapt this technology as a preliminary screening tool to identify patients at high risk of urosepsis and help guide targeted examinations or interventions.4
This retrospective study included patients with upper urinary tract calculi admitted to one Hospital between January 2016 and January 2020, and used imaging results, including urinary system ultrasound, excretory urogram, or abdominopelvic computed tomography as well as the participants’ complete medical history.4
To develop the model, the team applied a neural network consisting of three layers: an input layer that receives information, a hidden layer that processes information, and an output layer that calculates results. The network was then run with significant predictors as input variables and urosepsis risk as the output variable. The number of neurons in the input layer was the total number of covariables, while the output variable was dichotomous.4
For the model construction and validation, the participants were randomized into a training set with 1,214 patients, consisting of 135 cases and 1,079 controls. The validation set was comprised of 502 patients with 51 cases and 451 controls. Univariable and multivariable logistic regression analyses were performed to evaluate variables associated with urosepsis and to generate the artificial neural network model for the training set.4
The model revealed high concordance between the predicted and observed probabilities and a good agreement between the predicted and observed outcomes, for the training and validation sets. Based on ultrasound and urinalysis, the model showed encouraging outcomes regarding its ability to identify urosepsis in patients with upper urinary tract calculi.4
The results also show that the artificial neural network model had a better performance compared to the Nomogram model, which was used to predict the probability of patients with ureteral calculi developing into urosepsis in a previous study by Hu and colleagues.4
So, to summarize, over the last decade AI has become increasingly integrated into medicine and in urology. In this field, AI is being tested and implemented as a tool to aid in the diagnosis and treatment of different forms of cancer and other pathologies. AI-driven techniques are highly appealing because they can quickly analyze large amounts of data, such as medical images and tissue samples, to identify patterns and predict the likelihood of cancer. In addition, AI-based techniques show the potential to increase the accuracy of prostate cancer diagnosis and improve treatment plans for patients.3
There is, however, a lingering need to standardize outcome measures to help compare treatment modalities and make individualized patient-centered decisions. This will also help compare newer minimally invasive surgical therapies.1
Several limitations to the application of AI in medicine must also be considered. For example, AI systems rely on the quality and quantity of the data on which they are trained. Thus, they may not perform well when applied to real-world situations that differ from the initial data used to develop and train these algorithms. Additionally, there are concerns about the ethical implications of using AI in medical decision-making and the potential for bias in the algorithms that drive these systems.3
Thanks for joining us on this episode of Health Connect. Don't miss out on our next episode.
Discover more medical news on Viatris Connect.
Referências:
- 1. Shah M, Naik N, Hameed BZ, Paul R, Shetty DK, Ibrahim S, et al. Current Applications of Artificial Intelligence in Benign Prostatic Hyperplasia. Turk J Urol. 2022;48(4):262–267. Available at: http://dx.doi.org/10.5152/tud.2022.22028
- 2. Eminaga O, Shkolyar E, Breil B, Semjonow A, Boegemann M, Xing L, et al. Artificial intelligence-based prognostic model for urologic cancers: A SEER-based study. Cancers (Basel). 2022;14(13):3135. Available at: http://dx.doi.org/10.3390/cancers14133135
- 3. Chu TN, Wong EY, Ma R, Yang CH, Dalieh IS, Hung AJ. Exploring the use of artificial intelligence in the management of prostate cancer. Curr Urol Rep. 2023;24(5):231–240. Available at: Exploring the use of artificial intelligence in the management of prostate cancer
- 4. Hong X, Liu G, Chi Z, Yang T, Zhang Y. Predictive model for urosepsis in patients with Upper Urinary Tract Calculi based on ultrasonography and urinalysis using artificial intelligence learning. Int Braz J Urol. 2023;49(2):221–232. Available at: http://dx.doi.org/10.1590/S1677-5538.IBJU.2022.0450
Os links para todos os sites de terceiros são oferecidos como um serviço aos nossos visitantes e não implicam endosso, indicação ou recomendação do Health Connect. Os artigos vinculados são fornecidos apenas para fins informativos e não visam implicar uma atribuição pelo autor e/ou editor. O Health Connect se isenta de qualquer responsabilidade pelo conteúdo ou pelos serviços de outros sites. Recomendamos que você analise as políticas e condições de todos os sites que escolher acessar.
NON-2023-10037
NON-2023-2512