Computer-aided diagnosis (CAD) of prostate cancer on multiparametric magnetic resonance imaging (mpMRI), using artificial intelligence (AI), may reduce missed cancers and unnecessary biopsies, increase inter-observer agreement between radiologists, and alleviate pressures caused by rising case incidence and a shortage of specialist radiologists to read prostate mpMRI. However, well-designed evaluation studies are required to prove efficacy above current clinical practice. A systematic search of the MEDLINE, EMBASE, and arXiv electronic databases was conducted for studies that compared CAD for prostate cancer detection or classification on MRI against radiologist interpretation and a histopathological reference standard, in treatment-naïve men with a clinical suspicion of prostate cancer. Twenty-seven studies were included in the final analysis. Due to substantial heterogeneities in the included studies, a narrative synthesis is presented. Several studies reported superior diagnostic accuracy for CAD over radiologist interpretation on small, internal patient datasets, though this was not observed in the few studies that performed evaluation using external patient data. Our review found insufficient evidence to suggest the clinical deployment of artificial intelligence algorithms at present. Further work is needed to develop and enforce methodological standards, promote access to large diverse datasets, and conduct prospective evaluations before clinical adoption can be considered.

Background: Medical schools differ, particularly in their teaching, but it is unclear whether such differences matter, although influential claims are often made. The Medical School Differences (MedDifs) study brings together a wide range of measures of UK medical schools, including postgraduate performance, fitness to practise issues, specialty choice, preparedness, satisfaction, teaching styles, entry criteria and institutional factors.Method: Aggregated data were collected for 50 measures across 29 UK medical schools. Data includeinstitutional history(e.g. rate of production of hospital and GP specialists in the past),curricular influences(e.g. PBL schools, spend per student, staff-student ratio), selection measures(e.g. entry grades),teaching and assessment(e.g. traditional vs PBL, specialty teaching, self-regulated learning), student satisfaction, Foundation selection scores,Foundation satisfaction,postgraduate examination performance andfitness to practise(postgraduate progression, GMC sanctions). Six specialties (General Practice, Psychiatry, Anaesthetics, Obstetrics and Gynaecology, Internal Medicine, Surgery) were examined in more detail.Results: Medical school differences are stable across time (median alpha = 0.835). The 50 measures were highly correlated, 395 (32.2%) of 1225 correlations being significant withp < 0.05, and 201 (16.4%) reached a Tukey-adjusted criterion ofp < 0.0025.Problem-based learning (PBL) schools differ on many measures, including lower performance on postgraduate assessments. While these are in part explained by lower entry grades, a surprising finding is that schools such as PBL schools which reportedgreaterstudent satisfaction with feedback also showedlowerperformance at postgraduate examinations.More medical school teaching of psychiatry, surgery and anaesthetics did not result in more specialist trainees. Schools that taught more general practice did have more graduates entering GP training, but those graduates performed less well in MRCGP examinations, the negative correlation resulting from numbers of GP trainees and exam outcomes being affected both by non-traditional teaching and by greater historical production of GPs.Postgraduate exam outcomes were also higher in schools with more self-regulated learning, but lower in larger medical schools.A path model for 29 measures found a complex causal nexus, most measures causing or being caused by other measures. Postgraduate exam performance was influenced by earlier attainment, at entry to Foundation and entry to medical school (the so-called academic backbone), and by self-regulated learning.Foundation measures of satisfaction, including preparedness, had no subsequent influence on outcomes. Fitness to practise issues were more frequent in schools producing more male graduates and more GPs.Conclusions: Medical schools differ in large numbers of ways that are causally interconnected. Differences between schools in postgraduate examination performance, training problems and GMC sanctions have important implications for the quality of patient care and patient safety.

Background: What subjects UK medical schools teach, what ways they teach subjects, and how much they teach those subjects is unclear. Whether teaching differences matter is a separate, important question. This study provides a detailed picture of timetabled undergraduate teaching activity at 25 UK medical schools, particularly in relation to problem-based learning (PBL). Method: The Analysis of Teaching of Medical Schools (AToMS) survey used detailed timetables provided by 25 schools with standard 5-year courses. Timetabled teaching events were coded in terms of course year, duration, teaching format, and teaching content. Ten schools used PBL. Teaching times from timetables were validated against two other studies that had assessed GP teaching and lecture, seminar, and tutorial times. Results: A total of 47,258 timetabled teaching events in the academic year 2014/2015 were analysed, including SSCs (student-selected components) and elective studies. A typical UK medical student receives 3960 timetabled hours of teaching during their 5-year course. There was a clear difference between the initial 2 years which mostly contained basic medical science content and the later 3 years which mostly consisted of clinical teaching, although some clinical teaching occurs in the first 2 years. Medical schools differed in duration, format, and content of teaching. Two main factors underlay most of the variation between schools, Traditional vs PBL teaching and Structured vs Unstructured teaching. A curriculum map comparing medical schools was constructed using those factors. PBL schools differed on a number of measures, having more PBL teaching time, fewer lectures, more GP teaching, less surgery, less formal teaching of basic science, and more sessions with unspecified content. Discussion: UK medical schools differ in both format and content of teaching. PBL and non-PBL schools clearly differ, albeit with substantial variation within groups, and overlap in the middle. The important question of whether differences in teaching matter in terms of outcomes is analysed in a companion study (MedDifs) which examines how teaching differences relate to university infrastructure, entry requirements, student perceptions, and outcomes in Foundation Programme and postgraduate training.

Background: The diagnostic accuracy of diffusion-weighted imaging (DWI) to detect prostate cancer is well-established. DWI provides visual as well as quantitative means of detecting tumor, the apparent diffusion coefficient (ADC). Recently higher b-values have been used to improve DWI’s diagnostic performance. Purpose: To determine the diagnostic performance of high b-value DWI at detecting prostate cancer and whether quantifying ADC improves accuracy. Material and Methods: A comprehensive literature search of published and unpublished databases was performed. Eligible studies had histopathologically proven prostate cancer, DWI sequences using b-values ≥ 1000 s/mm², less than ten patients, and data for creating a 2 × 2 table. Study quality was assessed with QUADAS-2 (Quality Assessment of diagnostic Accuracy Studies). Sensitivity and specificity were calculated and tests for statistical heterogeneity and threshold effect performed. Results were plotted on a summary receiver operating characteristic curve (sROC) and the area under the curve (AUC) determined the diagnostic performance of high b-value DWI. Results: Ten studies met eligibility criteria with 13 subsets of data available for analysis, including 522 patients. Pooled sensitivity and specificity were 0.59 (95% confidence interval [CI], 0.57–0.61) and 0.92 (95% CI, 0.91–0.92), respectively, and the sROC AUC was 0.92. Subgroup analysis showed a statistically significant ( P = 0.03) improvement in accuracy when using tumor visual assessment rather than ADC. Conclusion: High b-value DWI gives good diagnostic performance for prostate cancer detection and visual assessment of tumor diffusion is significantly more accurate than ROI measurements of ADC.

False positive multiparametric magnetic resonance imaging (mpMRI) phenotypes prompt unnecessary biopsies. The Prostate MRI Imaging Study (PROMIS) provides a unique opportunity to explore such phenotypes in biopsy-naïve men with raised prostate-specific antigen (PSA) and suspected cancer. To compare mpMRI lesions in men with/without significant cancer on transperineal mapping biopsy (TPM). PROMIS participants (n=235) underwent mpMRI followed by a combined biopsy procedure at University College London Hospital, including 5-mm TPM as the reference standard. Patients were divided into four mutually exclusive groups according to TPM findings: (1) no cancer, (2) insignificant cancer, (3) definition 2 significant cancer (Gleason ≥3+4 of any length and/or maximum cancer core length ≥4mm of any grade), and (4) definition 1 significant cancer (Gleason ≥4+3 of any length and/or maximum cancer core length ≥6mm of any grade). Index and/or additional lesions present in 178 participants were compared between TPM groups in terms of number, conspicuity, volume, location, and radiological characteristics. Most lesions were located in the peripheral zone. More men with significant cancer had two or more lesions than those without significant disease (67% vs 37%; p< 0.001). In the former group, index lesions were larger (mean volume 0.68 vs 0.50 ml; p< 0.001, Wilcoxon test), more conspicuous (Likert 4-5: 79% vs 22%; p< 0.001), and diffusion restricted (mean apparent diffusion coefficient [ADC]: 0.73 vs 0.86; p< 0.001, Wilcoxon test). In men with Likert 3 index lesions, log2PSA density and index lesion ADC were significant predictors of definition 1/2 disease in a logistic regression model (mean cross-validated area under the receiver-operator characteristic curve: 0.77 [95% confidence interval: 0.67-0.87]). Significant cancer-associated MRI lesions in biopsy-naïve men have clinical-radiological differences, with lesions seen in prostates without significant disease. MRI-calculated PSA density and ADC could predict significant cancer in those with indeterminate MRI phenotypes. Magnetic resonance imaging (MRI) lesions that mimic prostate cancer but are, in fact, benign prompt unnecessary biopsies in thousands of men with raised prostate-specific antigen. In this study we found that, on closer look, such false positive lesions have different features from cancerous ones. This means that doctors could potentially develop better tools to identify cancer on MRI and spare some patients from unnecessary biopsies.

In this work, we consider the task of pairwise cross-modality image registration, which may benefit from exploiting additional images available only at training time from an additional modality that is different to those being registered. As an example, we focus on aligning intra-subject multiparametric Magnetic Resonance (mpMR) images, between T2-weighted (T2w) scans and diffusion-weighted scans with high b-value (DWI $_{\text {high-b}}$ ). For the application of localising tumours in mpMR images, diffusion scans with zero b-value (DWI $_{\text {b}={0}}$ ) are considered easier to register to T2w due to the availability of corresponding features. We propose a learning from privileged modality algorithm, using a training-only imaging modality DWI $_{\text {b}={0}}$ , to support the challenging multi-modality registration problems. We present experimental results based on 369 sets of 3D multiparametric MRI images from 356 prostate cancer patients and report, with statistical significance, a lowered median target registration error of 4.34 mm, when registering the holdout DWI $_{\text {high-b}}$ and T2w image pairs, compared with that of 7.96 mm before registration. Results also show that the proposed learning-based registration networks enabled efficient registration with comparable or better accuracy, compared with a classical iterative algorithm and other tested learning-based methods with/without the additional modality. These compared algorithms also failed to produce any significantly improved alignment between DWI $_{\text {high-b}}$ and T2w in this challenging application.