Dataset Information

A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging.

ABSTRACT: Prostate magnetic resonance imaging has become the imaging standard for prostate cancer in various clinical settings, with interpretation standardized according to the Prostate Imaging Reporting and Data System (PI-RADS). Each year, hundreds of scientific studies that report on the diagnostic performance of PI-RADS are published. To keep up with this ever-increasing evidence base, systematic reviews and meta-analyses are essential. As systematic reviews are highly resource-intensive, we investigated whether a machine learning framework can reduce the manual workload and speed up the screening process (title and abstract). We used search results from a living systematic review of the diagnostic performance of PI-RADS (1585 studies, of which 482 were potentially eligible after screening). A naïve Bayesian classifier was implemented in an active learning environment for classification of the titles and abstracts. Our outcome variable was the percentage of studies that can be excluded after 95% of relevant studies have been identified by the classifier (work saved over sampling: WSS@95%). In simulation runs of the entire screening process (controlling for classifier initiation and the frequency of classifier updating), we obtained a WSS@95% value of 28% (standard error of the mean ±0.1%). Applied prospectively, our classification framework would translate into a significant reduction in manual screening effort.

Patient summary

Systematic reviews of scientific evidence are labor-intensive and take a lot of time. For example, many studies on prostate cancer diagnosis via MRI (magnetic resonance imaging) are published every year. We describe the use of machine learning to reduce the manual workload in screening search results. For a review of MRI for prostate cancer diagnosis, this approach reduced the screening workload by about 28%.

SUBMITTER: Nedelcu A

PROVIDER: S-EPMC10495598 | biostudies-literature | 2023 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging.

Nedelcu Andrea A Oerther Benedict B Engel Hannes H Sigle August A Schmucker Christine C Schoots Ivo G IG Eisenblätter Michel M Benndorf Matthias M

European urology open science 20230830

Prostate magnetic resonance imaging has become the imaging standard for prostate cancer in various clinical settings, with interpretation standardized according to the Prostate Imaging Reporting and Data System (PI-RADS). Each year, hundreds of scientific studies that report on the diagnostic performance of PI-RADS are published. To keep up with this ever-increasing evidence base, systematic reviews and meta-analyses are essential. As systematic reviews are highly resource-intensive, we investig ...[more]

PMID: 37705517

Similar Datasets

Project description:BackgroundThe use of intraprostatic gold fiducial markers (FMs) ensures highly accurate and precise image-guided radiation therapy for patients diagnosed with prostate cancer thanks to the ease of localising FMs on photon-based imaging, like Computed Tomography (CT) images. Recently, Magnetic Resonance (MR)-only radiotherapy has been proposed to simplify the workflow and reduce possible systematic uncertainties. A critical, determining factor in the accuracy of such an MR-only simulation will be accurate FM localisation using solely MR images.PurposeThe aim of this study is to evaluate the performances of manual MR-based FM localisation within a clinical environment.MethodsWe designed a study in which 5 clinically involved radiation therapy technicians (RTTs) independently localised the gold FMs implanted in 16 prostate cancer patients in two scenarios: employing a single MR sequence or a combination of sequences. Inter-observer precision and accuracy were assessed for the two scenarios for localisation in terms of 95% limit of agreement on single FMs (LoA)/ centre of mass (LoA CM) and inter-marker distances (IDs), respectively.ResultsThe number of precisely located FMs (LoA <2 mm) increased from 38/48 to 45/48 FMs when localisation was performed using multiple sequences instead of single one. When performing localisation on multiple sequences, imprecise localisation of the FMs (3/48 FMs) occurred for 1/3 implanted FMs in three different patients. In terms of precision, we obtained LoA CM within 0.25 mm in all directions over the precisely located FMs. In terms of accuracy, IDs difference of manual MR-based localisation versus CT-based localisation was on average (±1 STD) 0.6 ±0.6 mm.ConclusionsFor both the investigated scenarios, the results indicate that when FM classification was correct, the precision and accuracy are high and comparable to CT-based FM localisation. We found that use of multiple sequences led to better localisation performances compared with the use of single sequence. However, we observed that, due to the presence of calcification and motion, the risk of mislocated patient positioning is still too high to allow the sole use of manual FM localisation. Finally, strategies to possibly overcome the current challenges were proposed.

Project description:ObjectiveProstate lesions detected with multiparametric magnetic resonance imaging (mpMRI) are classified for their malignant potential according to the Prostate Imaging-Reporting And Data System (PI-RADS™2). In this study, we evaluate the diagnostic accuracy of the mpMRI with and without gadolinium, with emphasis on the added diagnostic value of the dynamic contrast enhancement (DCE).Materials and methodsThe study was retrospective for 286 prostate lesions / 213 eligible patients, n = 116/170, and 49/59% malignant for the peripheral (Pz) and transitional zone (Tz), respectively. A stereotactic MRI-guided prostate biopsy served as the histological ground truth. All patients received a mpMRI with DCE. The influence of DCE in the prediction of malignancy was analyzed by blinded assessment of the imaging protocol without DCE and the DCE separately.ResultsSignificant (CSPca) and insignificant (IPca) prostate cancers were evaluated separately to enhance the potential effects of the DCE in the detection of CSPca. The Receiver Operating Characteristics Area Under Curve (ROC-AUC), sensitivity (Se) and specificity (Spe) of PIRADS-without-DCE in the Pz was 0.70/0.47/0.86 for all cancers (IPca and CSPca merged) and 0.73/0.54/0.82 for CSPca. PIRADS-with-DCE for the same patients showed ROC-AUC/Se/Spe of 0.70/0.49/0.86 for all Pz cancers and 0.69/0.54/0.81 for CSPca in the Pz, respectively, p>0.05 chi-squared test. Similar results for the Tz, AUC/Se/Spe for PIRADS-without-DCE was 0.75/0.61/0.79 all cancers and 0.67/0.54/0.71 for CSPca, not influenced by DCE (0.66/0.47/0.81 for all Tz cancers and 0.61/0.39/0.75 for CSPca in Tz). The added Se and Spe of DCE for the detection of CSPca was 88/34% and 78/33% in the Pz and Tz, respectively.ConclusionDCE showed no significant added diagnostic value and lower specificity for the prediction of CSPca compared to the non-enhanced sequences. Our results support that gadolinium might be omitted without mitigating the diagnostic accuracy of the mpMRI for prostate cancer.

Project description:PurposeTo assess the diagnostic weight of sequence-specific magnetic resonance features in characterizing clinically significant prostate cancers (csPCa).Materials and methodsWe used a prospective database of 262 patients who underwent T2-weighted, diffusion-weighted, and dynamic contrast-enhanced (DCE) imaging before prostatectomy. For each lesion, two independent readers (R1, R2) prospectively defined nine features: shape, volume (V_Max), signal abnormality on each pulse sequence, number of pulse sequences with a marked (S_Max) and non-visible (S_Min) abnormality, likelihood of extracapsular extension (ECE) and PSA density (dPSA). Overall likelihood of malignancy was assessed using a 5-level Likert score. Features were evaluated using the area under the receiver operating characteristic curve (AUC). csPCa was defined as Gleason ≥7 cancer (csPCa-A), Gleason ≥7(4+3) cancer (csPCa-B) or Gleason ≥7 cancer with histological extraprostatic extension (csPCa-C).ResultsFor csPCa-A, the Signal1 model (S_Max+S_Min) provided the best combination of signal-related variables, for both readers. The performance was improved by adding V_Max, ECE and/or dPSA, but not shape. All models performed better with DCE findings than without. When moving from csPCa-A to csPCa-B and csPCa-C definitions, the added value of V_Max, dPSA and ECE increased as compared to signal-related variables, and the added value of DCE decreased. For R1, the best models were Signal1+ECE+dPSA (AUC = 0,805 [95%CI:0,757-0,866]), Signal1+V_Max+dPSA (AUC = 0.823 [95%CI:0.760-0.893]) and Signal1+ECE+dPSA [AUC = 0.840 (95%CI:0.774-0.907)] for csPCa-A, csPCA-B and csPCA-C respectively. The AUCs of the corresponding Likert scores were 0.844 [95%CI:0.806-0.877, p = 0.11], 0.841 [95%CI:0.799-0.876, p = 0.52]) and 0.849 [95%CI:0.811-0.884, p = 0.49], respectively. For R2, the best models were Signal1+V_Max+dPSA (AUC = 0,790 [95%CI:0,731-0,857]), Signal1+V_Max (AUC = 0.813 [95%CI:0.746-0.882]) and Signal1+ECE+V_Max (AUC = 0.843 [95%CI: 0.781-0.907]) for csPCa-A, csPCA-B and csPCA-C respectively. The AUCs of the corresponding Likert scores were 0. 829 [95%CI:0.791-0.868, p = 0.13], 0.790 [95%CI:0.742-0.841, p = 0.12]) and 0.808 [95%CI:0.764-0.845, p = 0.006]), respectively.ConclusionCombination of simple variables can match the Likert score's results. The optimal combination depends on the definition of csPCa.

Dataset Information

A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging.

Patient summary

Publications

A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets