In this article, we present a methodological framework that meets novel requirements emerging from upcoming types of accelerated and highly configurable neuromorphic hardware systems. We describe in detail a device with 45 million programmable and dynamic synapses that is currently under development, and we sketch the conceptual challenges that arise from taking this platform into operation. More specifically, we aim at the establishment of this neuromorphic system as a flexible and neuroscientifically valuable modeling tool that can be used by non-hardware experts. We consider various functional aspects to be crucial for this purpose, and we introduce a consistent workflow with detailed descriptions of all involved modules that implement the suggested steps: The integration of the hardware interface into the simulator-independent model description language PyNN; a fully automated translation between the PyNN domain and appropriate hardware configurations; an executable specification of the future neuromorphic system that can be seamlessly integrated into this biology-to-hardware mapping process as a test bench for all software layers and possible hardware design modifications; an evaluation scheme that deploys models from a dedicated benchmark library, compares the results generated by virtual or prototype hardware devices with reference software simulations and analyzes the differences. The integration of these components into one hardware-software workflow provides an ecosystem for ongoing preparative studies that support the hardware design process and represents the basis for the maturity of the model-to-hardware mapping software. The functionality and flexibility of the latter is proven with a variety of experimental results.

The evaluation of bone marrow morphology by experienced hematopathologists is essential in the diagnosis of acute myeloid leukemia (AML); however, it suffers from a lack of standardization and inter-observer variability. Deep learning (DL) can process medical image data and provides data-driven class predictions. Here, we apply a multi-step DL approach to automatically segment cells from bone marrow images, distinguish between AML samples and healthy controls with an area under the receiver operating characteristic (AUROC) of 0.9699, and predict the mutation status of Nucleophosmin 1 (NPM1)—one of the most common mutations in AML—with an AUROC of 0.92 using only image data from bone marrow smears. Utilizing occlusion sensitivity maps, we observed so far unreported morphologic cell features such as a pattern of condensed chromatin and perinuclear lightening zones in myeloblasts of NPM1-mutated AML and prominent nucleoli in wild-type NPM1 AML enabling the DL model to provide accurate class predictions.

Machine learning (ML) is rapidly emerging in several fields of cancer research. ML algorithms can deal with vast amounts of medical data and provide a better understanding of malignant disease. Its ability to process information from different diagnostic modalities and functions to predict prognosis and suggest therapeutic strategies indicates that ML is a promising tool for the future management of hematologic malignancies; acute myeloid leukemia (AML) is a model disease of various recent studies. An integration of these ML techniques into various applications in AML management can assure fast and accurate diagnosis as well as precise risk stratification and optimal therapy. Nevertheless, these techniques come with various pitfalls and need a strict regulatory framework to ensure safe use of ML. This comprehensive review highlights and discusses recent advances in ML techniques in the management of AML as a model disease of hematologic neoplasms, enabling researchers and clinicians alike to critically evaluate this upcoming, potentially practice-changing technology.

Precision oncology is grounded in the increasing understanding of genetic and molecular mechanisms that underly malignant disease and offer different treatment pathways for the individual patient. The growing complexity of medical data has led to the implementation of machine learning techniques that are vastly applied for risk assessment and outcome prediction using either supervised or unsupervised learning. Still largely overlooked is reinforcement learning (RL) that addresses sequential tasks by exploring the underlying dynamics of an environment and shaping it by taking actions in order to maximize cumulative rewards over time, thereby achieving optimal long-term outcomes. Recent breakthroughs in RL demonstrated remarkable results in gameplay and autonomous driving, often achieving human-like or even superhuman performance. While this type of machine learning holds the potential to become a helpful decision support tool, it comes with a set of distinctive challenges that need to be addressed to ensure applicability, validity and safety. In this review, we highlight recent advances of RL focusing on studies in oncology and point out current challenges and pitfalls that need to be accounted for in future studies in order to successfully develop RL-based decision support systems for precision oncology.

Background Acute promyelocytic leukemia (APL) is considered a hematologic emergency due to high risk of bleeding and fatal hemorrhages being a major cause of death. Despite lower death rates reported from clinical trials, patient registry data suggest an early death rate of 20%, especially for elderly and frail patients. Therefore, reliable diagnosis is required as treatment with differentiation-inducing agents leads to cure in the majority of patients. However, diagnosis commonly relies on cytomorphology and genetic confirmation of the pathognomonic t(15;17). Yet, the latter is more time consuming and in some regions unavailable. Methods In recent years, deep learning (DL) has been evaluated for medical image recognition showing outstanding capabilities in analyzing large amounts of image data and provides reliable classification results. We developed a multi-stage DL platform that automatically reads images of bone marrow smears, accurately segments cells, and subsequently predicts APL using image data only. We retrospectively identified 51 APL patients from previous multicenter trials and compared them to 1048 non-APL acute myeloid leukemia (AML) patients and 236 healthy bone marrow donor samples, respectively. Results Our DL platform segments bone marrow cells with a mean average precision and a mean average recall of both 0.97. Further, it achieves high accuracy in detecting APL by distinguishing between APL and non-APL AML as well as APL and healthy donors with an area under the receiver operating characteristic of 0.8575 and 0.9585, respectively, using visual image data only. Conclusions Our study underlines not only the feasibility of DL to detect distinct morphologies that accompany a cytogenetic aberration like t(15;17) in APL, but also shows the capability of DL to abstract information from a small medical data set, i. e. 51 APL patients, and infer correct predictions. This demonstrates the suitability of DL to assist in the diagnosis of rare cancer entities. As our DL platform predicts APL from bone marrow smear images alone, this may be used to diagnose APL in regions were molecular or cytogenetic subtyping is not routinely available and raise attention to suspected cases of APL for expert evaluation.

Achievement of complete remission (CR) signifies a crucial milestone in the therapy of acute myeloid leukemia (AML) while refractory disease is associated with dismal outcomes. Hence, accurately identifying patients at risk is essential to individually tailor treatment concepts to disease biology. We used nine machine learning (ML) models to predict CR and 2-year overall survival (OS) in a large multi-center cohort of 1383 AML patients who received intensive induction therapy using clinical, laboratory, cytogenetic and molecular genetic data and validated our results on an external multicenter cohort. Our ML models autonomously selected predictive features both including established markers of favorable or adverse risk as well as identifying markers of so-far controversial relevance. De novo AML, extramedullary AML, double-mutated (dm) CEBPA, mutations of CEBPA-bZIP, NPM1, FLT3-ITD, ASXL1, RUNX1, SF3B1, IKZF1, TP53, U2AF1, t(8;21), inv(16)/t(16;16), del(5)/del(5q), del(17)/del(17p), normal or complex karyotypes, age and hemoglobin at initial diagnosis were statistically significant markers predictive of CR, while t(8;21), del(5)/del(5q), inv(16)/t(16;16), del(17)/del(17p), dmCEBPA, CEBPA-bZIP, NPM1, FLT3-ITD, DNMT3A, SF3B1, U2AF1, TP53, age, white blood cell count, peripheral blast count, serum LDH and Hb at initial diagnosis as well as extramedullary manifestations were predictive for 2-year OS. For prediction of CR and 2-year OS, AUROCs ranged between 0.77 – 0.86 and 0.63 and 0.74, respectively in our test set and 0.71 – 0.80 and 0.65 – 0.75 in the external validation cohort. We demonstrate the feasibility of ML for risk stratification in AML as a model disease for hematologic neoplasms using a scalable and reusable ML framework. Our study illustrates the clinical applicability of ML as a decision support system in hematology.

Achievement of complete remission (CR) signifies a crucial milestone in the therapy of acute myeloid leukemia (AML) while refractory disease is associated with dismal outcomes. Hence, accurately identifying patients at risk is essential to tailor treatment concepts individually to disease biology. Machine Learning (ML) is a branch of computer science that can process large data sets for a plethora of purposes. The underlying mechanism does not necessarily begin with a manually drafted hypothesis model. Rather the ML algorithms can detect patterns in pre-processed data and derive abstract information. We used ML to predict CR and 2-year overall survival (OS) in a large multi-center cohort of 1383 AML patients who received intensive induction therapy using clinical, laboratory, cytogenetic and molecular genetic data. To enable a customizable and reusable technological approach and achieve optimal results, we designed a data-driven platform with an embedded, automated ML pipeline integrating state-of-the-art software technology for data management and ML models. The platform consists of five scalable modules for data import and modelling, data transformation, model refinement, machine learning algorithms, feature support and performance feedback that are executed in an iterative manner to approach step-wisely the optimal configuration. To reduce dimensionality and the the risk of overfitting, dynamic feature selection was used, i.e. features were selected according to their support by feature selection algorithms. To be included in an ML model, a feature had to pass a pre-determined threshold of overall predictive power determined by summing the normalized scores of the feature selection algorithms. Features below the threshold were automatically excluded from the ML models for the respective iteration. In that way, features of high redundancy or low entropy were automatically filtered out. Our classification algorithms were completely agnostic of pre-existing risk classifications and autonomously selected predictive features both including established markers of favorable or adverse risk as well as identifying markers of so-far controversial relevance. De novo AML, extramedullary AML, double-mutated (dm) CEBPA, mutations of CEBPA-bZIP, NPM1, FLT3-ITD, ASXL1, RUNX1, SF3B1, IKZF1, TP53, U2AF1, t(8;21), inv(16)/t(16;16), del5/del5q, del17, normal or complex karyotypes, age and hemoglobin at initial diagnosis were statistically significant markers predictive of CR while t(8;21), del5/del5q, inv(16)/t(16;16), del17, dm CEBPA, CEBPA-bZIP, NPM1, FLT3-ITD , DNMT3A, SF3B1, U2AF1, TP53, age, white blood cell count, peripheral blast count, serum LDH and Hb at initial diagnosis as well as extramedullary manifestations were predictive for 2-year OS. For prediction of CR and 2-year OS, AUROCs ranged between 0.77 - 0.86 and 0.63 - 0.74, respectively. We provide a method to automatically select predictive features from different data types, cope with gaps and redundancies, apply and optimize different ML models, and evaluate optimal configurations in a scalable and reusable ML platform. In a proof-of-concept manner, our algorithms utilize both established markers of favorable or adverse risk and also provide further evidence for the roles of U2AF1, IKZF1, SF3B1, DNMT3A and bZIP mutations of CEBPA in AML risk prediction. Our study serves as a fundament for prospective validation and data-driven ML-guided risk assessment in AML at initial diagnosis for the individual patient. Image caption: Patient features were automatically selected by machine learning to predict complete remission (CR) and 2-year overall survival (OS) after intensive induction therapy. Based on a continuous feature support metric with a predefined cut-off of 0.5 (determined by optimal classification performance), 27 and 25 features were automatically selected for prediction of CR (A) and 2-year OS (C), respectively. For each of these features predicted by machine learning, odds ratios and 95% confidence intervals (CI) were calculated for CR (B) and 2 year OS (D). BMB: bone marrow blast count; FLT3h/low: FLT3-ITD ratio, h=high>0.5; Hb: hemoglobin; karyotype, c: complex aberrant karyotype (≥ 3 aberrations); karyotype, n: normal karyotype (no aberrations); LDH: lactate dehydrogenase; PBB: peripheral blood blast count; PLT: platelet count; WBC: white blood cell count. Figure 1 Figure 1. Disclosures Schetelig: Roche: Honoraria, Other: lecture fees; Novartis: Honoraria, Other: lecture fees; BMS: Honoraria, Other: lecture fees; Abbvie: Honoraria, Other: lecture fees; AstraZeneca: Honoraria, Other: lecture fees; Gilead: Honoraria, Other: lecture fees; Janssen: Honoraria, Other: lecture fees . Platzbecker: Janssen: Honoraria; Celgene/BMS: Honoraria; AbbVie: Honoraria; Novartis: Honoraria; Takeda: Honoraria; Geron: Honoraria. Müller-Tidow: Pfizer: Research Funding; Janssen: Consultancy, Research Funding; Bioline: Research Funding. Baldus: Celgene/BMS: Honoraria; Amgen: Honoraria; Novartis: Honoraria; Jazz: Honoraria. Krause: Siemens: Research Funding; Takeda: Honoraria; Pfizer: Honoraria; art-tempi: Honoraria; Kosmas: Honoraria; Gilead: Other: travel support; Abbvie: Other: travel support. Haenel: Bayer Vital: Honoraria; Jazz: Consultancy, Honoraria; GSK: Consultancy; Takeda: Consultancy, Honoraria; Novartis: Consultancy, Honoraria; Roche: Consultancy, Honoraria; Amgen: Consultancy; Celgene: Consultancy, Honoraria. Schliemann: Philogen S.p.A.: Consultancy, Honoraria, Research Funding; Abbvie: Consultancy, Other: travel grants; Astellas: Consultancy; AstraZeneca: Consultancy; Boehringer-Ingelheim: Research Funding; BMS: Consultancy, Other: travel grants; Jazz Pharmaceuticals: Consultancy, Research Funding; Novartis: Consultancy; Roche: Consultancy; Pfizer: Consultancy. Middeke: Roche: Consultancy, Honoraria; Janssen: Consultancy, Honoraria, Research Funding; Abbvie: Consultancy, Honoraria; Pfizer: Consultancy, Honoraria; Jazz: Consultancy; Astellas: Consultancy, Honoraria; Sanofi: Honoraria, Research Funding; Novartis: Consultancy; Gilead: Consultancy; Glycostem: Consultancy; UCB: Honoraria.

The clinical monitoring of walking generates enormous amounts of data that contain extremely valuable information. Therefore, machine learning (ML) has rapidly entered the research arena to analyze and make predictions from large heterogeneous datasets. Such data-driven ML-based applications for various domains become increasingly applicable, and thus their software qualities are taken into focus. This work provides a proof of concept for applying state-of-the-art ML technology to predict the distance travelled of the 2-min walk test, an important neurological measurement which is an indicator of walking endurance. A transparent lean approach was emphasized to optimize the results in an explainable way and simultaneously meet the specified software requirements for a generic approach. It is a general-purpose strategy as a fractional–factorial design benchmark combined with standardized quality metrics based on a minimal technology build and a resulting optimized software prototype. Based on 400 training and 100 validation data, the achieved prediction yielded a relative error of 6.1% distributed over multiple experiments with an optimized configuration. The Adadelta algorithm (LR=0.000814, fModelSpread=5, nModelDepth=6, nepoch=1000) performed as the best model, with 90% of the predictions with an absolute error of <15 m. Factors such as gender, age, disease duration, or use of walking aids showed no effect on the relative error. For multiple sclerosis patients with high walking impairment (EDSS Ambulation Score ≥6), the relative difference was significant (n=30; 24.0%; p<0.050). The results show that it is possible to create a transparently working ML prototype for a given medical use case while meeting certain software qualities.